CHANGELOG.md

CHANGELOG

All notable changes to XYPH will be documented in this file.

Unreleased

Added

• search CLI command — unified quest search with keyword matching (title, ID, description), --status filtering (including GRAVEYARD), and --stats mode for counts by status. Always includes graveyard quests so the full graph is searchable. Usage: xyph search "MCP", xyph search --status GRAVEYARD, xyph search --stats.
• 9 invariants seeded into the WARP graph as content-attached nodes — graph-is-truth, principal-general-authority, deterministic-convergence, immutable-provenance, authorized-intent, substrate-boundary, policy-is-plastic, frontier-not-assignment, witness-before-done. Each carries the full METHOD structure (what must remain true / why it matters / how to check) and is readable from any branch via xyph show invariant:*.
• 5 legends (WARP, GOV, SURF, PROV, FLOW) with protects edges to invariants — eternal domains that span the project lifetime, replacing temporal campaign-based grouping. Each legend defines its sponsored human/agent roles, hills, and verification criteria.
• First bearing set as a graph-native signpost — bearing:current captures direction, last shipped work, what's next, and what feels wrong. Updated at cycle boundaries.
• 6 new schema prefixes (invariant, legend, cycle, design, retro, bearing) and 3 new edge types (protects, contains, closes) preparing the graph ontology for METHOD-aligned cycle discipline.

Changed

• Backlog triaged from 158 open quests to 53 — 11 quests sealed as already completed, ~88 graveyarded with full "WHAT IT WAS" intent reconstructions, 24 stubs merged into 6 consolidated cycle-worthy items, 18 rethought (12 graveyarded, 1 reset, 5 reframed with new titles and descriptions reflecting METHOD adoption).
• Doctor no longer flags BACKLOG, GRAVEYARD, or DONE quests for readiness and governed-completion gaps — these scans now only check PLANNED/READY/IN_PROGRESS quests, eliminating 52 false diagnostics from graveyarded and sealed work. Doctor diagnostics went from 55 to 3 (all legitimate PLANNED quest work).
• DoctorService test fixture updated — governed-quest-incomplete test now uses PLANNED status instead of BACKLOG, correctly reflecting that BACKLOG quests should not trigger governance checks.

Fixed

• Replaced rg with grep in diagnostic logging acceptance test — the cycle 0022 test spawned rg (ripgrep) directly, which isn't guaranteed on CI runners. When missing, spawnSync returned undefined stdout causing a TypeError.

Changed

• README rewritten from the ground up — the README was a changelog masquerading as onboarding. It has been rewritten to lead with what XYPH is, move straight into a walkthrough, and present the full CLI/TUI/API surface grounded in what actually exists today. The 276-line worldlines substrate essay has been extracted to docs/WORLDLINES.md. The CLI reference table now covers all 61 commands across 18 groups. The milestone table reflects current reality. The TUI section describes what a user actually sees, not accumulated patch-note deltas.

• The design corpus now captures the git-warp / XYPH substrate boundary more explicitly — design/substrate-alignment.md now spells out the target model in which git-warp owns worldlines, observer-relative reads, working sets, speculative ticking, transfer/collapse primitives, and substrate receipts while XYPH owns policy, governance, cases, and product meaning. The invariants and product model now mirror that split, and a new cycle note, design/cycles/0004-substrate-alignment.md, captures the first bounded pivot slice so future work stops expanding XYPH's omnibus graph reconstruction by default.

• Governed cases now have a human page, durable decisions, and linked follow-on work — suggestion pages can now drill into an attached governed case, the new case page shows the decision question, subject scope, attached briefs, and a receipt-like decision summary, and pressing d records a durable human decision that stays distinct from briefs while compiling follow-on work into existing quest or proposal primitives by default.

• Cycle methodology now treats playbacks as a formal hill-measurement gate — the contributing guide and design corpus now make the four internal checkpoints explicit (doctrine, spec, semantic, surface), allow local red iteration while requiring pushed work to be green, and define formal human and agent playbacks as the place where sponsor-actor outcomes and hills are actually judged.

• The design corpus now names the larger governed product loop XYPH is meant to embody — design/product-loop.md now captures XYPH not just as a TUI or agent-native CLI, but as the governed conduit through which project shape is observed, suggested, elevated, prepared for judgment, decided, applied, verified, and reconciled. The document now scopes that loop to shape-changing decisions rather than routine execution, requires cases to carry a real decision question and subject scope, treats off-graph observations as valid once reified with provenance and uncertainty, trims the durable ontology down to a minimal case -> brief -> decision -> linked follow-on work spine, keeps readiness/quorum/staleness as derived judgments until proven necessary, splits decision classification into orthogonal impact/risk/authority axes, prefers linked quests and existing execution primitives over minting fresh work nouns, adds decision receipts and non-linear transitions, and makes worldline-backed option sets part of serious judgment preparation instead of treating recommendations as prose only.

• Human and agent workflow doctrine is now explicit in the design corpus and contributing guide — the design corpus now says plainly that XYPH has one product with two operational lenses, that cycles should be designed from both perspectives, and that implementation should usually proceed through shared semantics and agent-native seams before the human page/TUI expression. The contributing guide now mirrors that rule as project methodology instead of leaving it implicit in chat.

• Alignment sweep now removes stale planning docs and retires obsolete TUI backlog items — the unified design corpus plus the graph backlog are now the only active planning surfaces. TASKS.md, docs/PLAN.md, and docs/TUI-plan.md are being removed instead of lingering as misleading secondary plans, the README milestone table now clearly reads as historical spine rather than active truth, and an alignment sweep record now captures the rationale. The sweep also retired obsolete backlog items that no longer fit the landing/page model or current product priorities, including task:tui-suggestion-tab, task:dashboard-suggestion-widget, task:tui-quest-modal, and task:warp-explorer-view.

• Suggestion adoption now chooses a real work shape — accepting an AI suggestion no longer always lowers into a generic governed proposal. Quest-shaped suggestions now default into real backlog quests with visible provenance, while other suggestions can still be adopted as governed proposals, and suggestion accept --as quest|proposal plus the suggestion-page adoption flow now make that choice explicit.

• Suggestion pages now support real adoption flows and [AI] explainability — AI suggestion pages no longer stop at passive inspection. Operators can now adopt a suggestion into governed work, dismiss it with rationale, or mark it superseded by another artifact directly from the page-local action model, and pressing e opens a dedicated [AI] explainability modal instead of relying only on static transparency prose.

• TUI launch now has a first-class npm script — npm run tui now launches xyph-dashboard.ts, so the interactive XYPH surface no longer depends on remembering the raw npx tsx entrypoint.

• Suggestions now have TUI subviews and an in-app Ask-AI composer — the Suggestions lane now cycles through Incoming, Queued, Adopted, and Dismissed subviews with v, so advisory items are no longer one flat mixed queue. The cockpit also now exposes n as an in-app Ask-AI composer, letting operators queue ask-AI jobs directly from the TUI or an item page instead of dropping to the CLI.

• Suggestions are now easier to discover and track in the TUI — empty suggestion subviews now explain where the suggestions actually are and prompt operators to cycle views, suggestion rows now surface linked governed cases directly in the worklist, current-observer suggestions sort ahead of unrelated ones, and the My Stuff drawer now includes active suggestions so your own outstanding AI work stays visible.

• Suggestions now include explicit ask-AI jobs and agent pickup queues — ask-ai is now a first-class AI suggestion kind, xyph ask-ai now queues explicit request artifacts for agent pickup, the Suggestions lane/page now distinguishes queued ask-AI jobs from ordinary advisory suggestions, and the agent surface now exposes those artifacts through suggestionQueue in briefing, suggestion candidates in next, and shared suggestion semantics in context.

• Security floor raised for flatted — the dependency override now pins flatted@3.4.2, clearing the high-severity prototype-pollution advisory from the ESLint cache stack without changing XYPH runtime behavior.

• Suggestions are now a first-class cockpit lane and page type — the landing shell now has seven lanes instead of six, with a dedicated Suggestions lane between Settlement and Campaigns. Selecting a suggestion and pressing Enter now opens a breadcrumbed suggestion page with [AI] labeling, lifecycle/progress, graph context, AI transparency copy, and page-local commenting instead of forcing advisory artifacts through the landing inspector.

• XYPH product design now lives in a first-class design/ corpus — the old monolithic docs/XYPH_PRODUCT_DESIGN.md path is now a compatibility pointer, while the canonical design source has been split into design/README.md, design/sponsor-actors.md, design/hills.md, design/playbacks.md, design/invariants.md, and design/product-model.md. This makes sponsor actors, hills, playbacks, and XYPH’s non-negotiable design invariants explicit top-level artifacts instead of long sections buried in one file.

• Agent-originated suggestions are now explicitly first-class in the product design — the unified design and agent protocol docs now state that explicit ask-AI jobs are one auditable request path, not the only lawful source of suggestions. Agents may also emit unsolicited advisory suggestions while working, as long as those suggestions are recorded as visible graph-native artifacts instead of silently mutating truth.

• Agent governance packets now expose routed follow-up instead of inspect-only placeholders — xyph context --json for comparison-artifact:*, collapse-proposal:*, and attestation:* targets now includes governance-specific recommendationRequests plus richer recommended actions such as human-bound attest, collapse_preview, and collapse_live candidates. xyph briefing --json and xyph next --json now surface the same governance follow-up semantics, and xyph act --json rejects those governance-only actions with explicit human-only-action envelopes instead of vague unsupported-action failures.

• Agent work packets now cover submissions and governance end-to-end — xyph context --json no longer stops at quest targets. It now emits typed work packets for submission:*, patchset:*, comparison-artifact:*, collapse-proposal:*, and attestation:* targets, including shared semantic judgments and recommended next actions. xyph briefing --json now includes a governanceQueue, xyph next --json can surface governance inspect work, and xyph act --json now carries the same shared submission semantics on review / merge refusal, dry-run, success, and partial-failure envelopes when XYPH can derive them honestly.

• Review pages are now real, and agent intake commands now speak the shared semantic packet too — Review lane submissions now open a dedicated breadcrumbed review page instead of forcing submission governance through the landing inspector. That page now renders lifecycle/progress, blocker and missing-evidence judgments, next lawful actions, and page-local comment / approve / request-changes actions from shared domain semantics. In parallel, xyph briefing --json and xyph next --json now carry compatible shared semantic packets for quest work and submission review candidates, joining xyph context --json so the first human and agent verticals now speak one vocabulary (requirements, acceptanceCriteria, evidenceSummary, blockingReasons, missingEvidence, nextLawfulActions, claimability, expectedActor, attentionState).

• XYPH now defines a shared primitive model for the next human and agent verticals — docs/XYPH_PRODUCT_DESIGN.md now explicitly models work primitives (requirements, acceptance criteria, evidence, quests, submissions), governance primitives (comparison-artifact, attestation, collapse-proposal, blockingReasons, nextLawfulActions, expectedActor, claimability), and agent-native packet semantics. The same doc now names Hill 3 (governance legibility) and Hill 6 (agent productivity without shell archaeology) as the next leverage-maximizing vertical slices, and the canonical agent/architecture docs now mirror those shared semantics.

• XYPH product design now covers the agent-native surface as a first-class product — docs/XYPH_PRODUCT_DESIGN.md no longer stops at the human cockpit. It now applies the same IBM-inspired design framing to sponsor agents, defines agent-oriented hills, unifies human and agent surfaces in one design source of truth, and adds product direction for suggestion queues, explicit ask-AI jobs, suggestion-backed quest creation, and a future live graph-feed mode.

• XYPH product design now defines AI suggestions and [AI] transparency — docs/XYPH_PRODUCT_DESIGN.md now treats AI-generated suggestions as first-class advisory artifacts for both human and agent consumption, defines suggestion families and lifecycle expectations, and adds a dedicated [AI] component inspired by Carbon's AI label so AI-driven content is visibly marked and explainable instead of blending into ordinary graph truth.

• XYPH now has an explicit product-design source of truth — XYPH now defines the human-facing app through docs/XYPH_PRODUCT_DESIGN.md, a new design document that adapts IBM Design Thinking to the XYPH human surface using sponsor users, hills, playbacks, landing-vs-page rules, a page taxonomy, an attention model, and a capability matrix. README, TUI planning, architecture, and active-direction docs now point at that document instead of competing to define the product independently.

• Quest item pages now expose real write actions instead of just context — drill-in quest pages now render an explicit action strip, support graph-native quest comments, expose reopen for graveyarded quests, and keep claim/promote/reject/review affordances visible in the page-local footer/help/palette so the landing cockpit can stay a triage surface instead of carrying every mutation affordance forever.

• XYPH now treats the cockpit as a landing page instead of a one-screen worldview — the TUI now keeps the lane rail/worklist/inspector shell as the landing surface, adds a sixth top-level Graveyard lane for rejected/retired work, and lets operators press Enter on quest-facing records to open a dedicated breadcrumbed item page. Esc / Backspace now return from that page to the landing cockpit instead of forcing all detail into the inspector forever.

• Review and Settlement badges now mean attention, not just recency — those two lanes no longer clear their hottest signal simply because you visited them once. XYPH now keeps persistent action-needed badges for open review work, stale/blocking governance artifacts, pending attestations, and ready-to-settle proposals, while the inspector explains the selected item with an explicit Why hot reason.

• Freshness is now visit-aware instead of click-to-clear busywork — row-level freshness markers now disappear as soon as you actually highlight a row, while Shift+S remains the explicit “mark the whole lane seen” control. XYPH now persists per-item seen state alongside lane watermarks so visiting one row no longer lies about the rest of the lane.

• Help is now a contextual modal and the footer has one control row — the top footer line is back to status/selection only, the bottom footer line is now the sole keyboard-control line, and ? opens a scrollable in-app help modal built from the same context/global controls instead of replacing the cockpit with a disconnected BIJOU help screen.

• Cockpit lanes now remember what is new to each observer — the dashboard now persists per-lane “last seen” watermarks in ~/.xyph/dashboard-state.json, shows lane-level freshness badges in the rail, and marks newly-seen worklist rows so operators can tell what changed since they last left a lane instead of treating all recent state as equally fresh.

• The cockpit now speaks mouse as well as keyboard — the dashboard runtime now enables BIJOU mouse events, lane rail entries are clickable, worklist rows are selectable by click, wheel scrolling now works in the worklist, inspector, quest-tree modal, and My Stuff drawer, clicking outside the quest-tree modal dismisses it, and clicking outside the open drawer closes it.

• Selected quests now open a graph-native tree modal — press t on a selected quest to open a lineage / dependency modal that shows sovereign intent lineage, campaign placement, submission/scroll settlement artifacts, upstream dependencies, and downstream dependents without leaving the cockpit.

• The Now lane now has a real recent-activity mode — press v on the Now lane to switch between the forward-looking action queue and a chronological “what changed, who did it” stream built from quest, review, decision, and governance-artifact facts. The lane rail count, worklist chrome, inspector detail, command palette, and footer hints now all reflect the active mode instead of hiding recent activity in the drawer.

• Cockpit pane heights now honor the terminal viewport in both wide and narrow layouts — header boxes are now measured instead of guessed, pane body heights are derived from the actual rendered chrome, and the stacked narrow-layout branch now uses a real constrained column layout. This prevents worklist and inspector cards from drawing underneath the footer when the viewport gets tight.

• The cockpit hero now anchors observer identity on the right — the top bar no longer treats surface, observer, and worldline as one mushy inline detail string. XYPH now renders product + active surface on the left and observer/worldline identity on the right so the operator always knows who is seeing what reality.

• The “My Stuff” drawer is now a real wrapped side sheet instead of a truncation trap — opening the drawer now claims substantially more screen width, wraps quest/submission/activity text on whitespace, and includes governance events in the recent-activity feed so the side sheet can answer more of “what just happened and who did it?” without clipping every line down to fit.

• The footer no longer splits controls across two competing lines — status lives on the upper footer line, controls live on the lower line, and the help modal now reflects that same source of truth so the keyboard story stops contradicting itself.

• Footer chrome now hard-clamps to the terminal width — the status and hint bars no longer rely on terminal soft-wrap. Oversized footer text now truncates with an ellipsis instead of visually wrapping into the cockpit area and making the pane bottoms look clipped by the footer.

• Scrollable cockpit panes now show explicit vertical scrollbars with interaction fade — the worklist and inspector lanes now render a real track with arrows and thumb when navigation wakes them, then fade back out once interaction stops. The rail now overlays the pane edge instead of reserving a dead gutter column, so content keeps its maximum width when the scrollbar is hidden.

• The cockpit frame now takes its accent from the active lane — Now, Plan, Review, Settlement, and Campaigns each now own a distinct lane color, and the hero border, pane headers, selected lane label, and focused worklist card all pick up that surface accent instead of sharing one global chrome color.

• Focused worklist cards now fill their full body instead of stopping at the last glyph — the selected-row background is now padded across the whole inner card width before BIJOU renders it, so the active card reads like a true block selection rather than a highlight that only hugs the occupied text.

• Cockpit worklist rows now read as mini-cards across every lane — non-selected rows no longer collapse into bare text lines. Every row now carries an explicit bordered three-line shape, while the focused row still gets the stronger primary border and gradient title treatment. This also fixes a row-width accounting bug that could let metadata overflow its actual content column on tighter panes.

• The cockpit inspector can now be toggled off — i now hides or restores the right-hand inspector pane, letting the center worklist reclaim the width when the operator wants to scan lanes without detail chrome.

• Cockpit worklists now render as stable contained-list queues instead of fragile tables — the center pane no longer relies on BIJOU’s dense navigableTable() rendering for long uneven rows. Each lane now uses a fixed-height queue row with stable selection, clearer hierarchy, and explicit paging, which removes the shifting/truncation artifacts that showed up most clearly in Review and Campaigns.

• TUI hierarchy and navigation were tightened after first-flight feedback — low-contrast muted text is now used much more sparingly, inspector status headers no longer leak BIJOU surface objects as [object Object], PgUp / PgDn now page the center worklist, and Shift+PgUp / Shift+PgDn scroll the inspector.

• The TUI was redesigned from the ground up as the XYPH cockpit landing shell — the old six-view dashboard shell has been replaced by a single operator landing surface with seven lanes (Now, Plan, Review, Settlement, Suggestions, Campaigns, Graveyard), a persistent inspector, a command palette, and contextual mutation hotkeys. The TUI now reads more like one control plane than six unrelated screens.

• TUI floor raised to BIJOU ^3.1.0 — XYPH now consumes the latest published @flyingrobots/bijou, @flyingrobots/bijou-node, and @flyingrobots/bijou-tui line. The theme bridge, shader path, landing canvas, and string/surface seams in the CLI renderers were updated for BIJOU v3 semantics instead of keeping the dashboard pinned to the old ^1.8.0 contract.

• explain now diagnoses governance artifacts directly — pointing xyph api explain at a durable comparison-artifact:*, collapse-proposal:*, or attestation:* now returns stable reason codes plus suggested next commands, including the crucial distinction that approving a collapse-proposal:* does not satisfy the live execution gate when the bound comparison-artifact:* still lacks approval.

• Substrate floor raised to @git-stunts/git-warp@^14.16.2 — XYPH now consumes the latest published git-warp patch, which in turn refreshes its git-cas floor to ^5.3.2 and keeps the app pinned to the actually released substrate line.

• query is now the first real governance read surface — the hidden admin command no longer falls through to not_implemented. It now exposes governance.worklist for actionable compare/collapse queues and governance.series for durable artifact-lane history.

• Persisted governance artifacts now have readable lifecycle semantics — observe(entity.detail) now computes XYPH governance detail for durable comparison-artifact:*, collapse-proposal:*, and attestation:* nodes, including freshness, attestation summary, and execution-gate state where applicable, instead of exposing only raw graph properties.

• Durable compare/collapse artifacts now form explicit lineage — repeated persisted comparison-artifact:* and collapse-proposal:* records in the same governance lane now carry a stable artifact_series_key, and newer artifacts link to older ones through supersedes so XYPH can report whether an artifact is current or already superseded.

• collapse_worldline now settles committed content-clearing plans too — governed live collapse no longer fails closed when git-warp transfer planning emits clear_node_content or clear_edge_content. Those ops now lower through the published substrate clear-content primitives in the shared mutation kernel.

• collapse_worldline is no longer preview-only — the command still defaults to dry-run preview, but now accepts dryRun: false for governed live settlement into worldline:live when approving attestations bind to a persisted comparison-artifact:*.

• Collapse execution now gates on persisted comparison facts — live collapse no longer treats collapse-proposal:* as the approval anchor. The current gate is one or more approving attestation:* nodes targeting the durable comparison-artifact:* returned by compare_worldlines persist:true, keeping governance attached to the factual comparison baseline.

• Collapse artifact identity is now mode-stable instead of approval-shaped — collapse-proposal artifact identity now varies by preview-vs-execution mode (dryRun) but no longer depends on which attestationIds were supplied, so governance approval metadata does not redefine the artifact itself.

• Runtime graph selection no longer hardcodes xyph-roadmap — the CLI, TUI, coordinator daemon, graph inspection helper, ref-sync scripts, and maintenance scripts now resolve the runtime graph through a shared bootstrap resolver instead of assuming one baked-in graph name.

• Graph bootstrap now defaults to xyph and fails loudly on ambiguity — XYPH now resolves graph bootstrap from local .xyph.json, then ~/.xyph/config, then defaults. If the target repo already contains multiple git-warp graphs, or only a single non-default graph, XYPH now rejects startup until graph.name is configured explicitly instead of silently picking or creating the wrong graph.

• Off-repo graph bootstrap no longer probes ref namespaces by default — when repoPath points outside the current working repo, XYPH now requires an explicit graph.name instead of scanning that repo’s refs/warp/*. Sidecar repos stay supported, but the bootstrap boundary is now explicit.

• Substrate floor raised to @git-stunts/git-warp@^14.16.0 — XYPH now consumes the published substrate release that adds first-class committed content-clear patch primitives, keeping governed collapse execution pinned to a real released substrate version.

• Worldline composition terminology stays braid-shaped in the public API — the active README and canonical control-plane docs continue to use braid_worldlines as the public composition verb for co-present worldline effects, explicitly rejecting branch/rebase vocabulary for this capability.

• Observation coordinates now make braid backing explicit — working-set-backed canonical derived-worldline reads now report the backing working-set ID plus overlay/braid details in observation.backing instead of only exposing a derived frontier digest.

• Braided conflict reads now warn on singleton self-erasure — observe(conflicts) now adds an explicit structural warning when braided overlays compete on a singleton LWW property winner, because that application modeling pattern cannot honestly represent co-presence under braid.

• Compare/collapse freshness is now operationally scoped — XYPH now computes compare/collapse freshness over an explicit governance-excluding node-prefix scope while still carrying the raw whole-graph substrate fact for audit. Durable governance chatter on worldline:live no longer self-invalidates operational comparison baselines.

Added

• xyph suggest records visible advisory suggestion artifacts — XYPH now has a dedicated CLI entry point for emitting [AI] suggestion records with explicit kind, audience, target, related entities, evidence, and recommended next action fields. Suggestions can be requested or spontaneous, but they are always recorded as visible graph-native advisory content instead of silently mutating plan truth.

• Settlement lane in the new cockpit shell — the redesigned TUI now gives persisted comparison-artifact:*, collapse-proposal:*, and attestation:* records a first-class Settlement lane inside the main cockpit, keeping compare/attest/collapse work on the same operator surface as quests, reviews, and campaigns.

• Governance artifacts in dashboard snapshots — GraphSnapshot now includes typed durable governance artifacts projected from the graph, so the TUI and future human-facing surfaces can consume the same freshness, attestation, and execution-state truth that already powers observe(entity.detail), query, and explain.

• Governed live collapse_worldline execution — xyph api collapse_worldline now executes settlement into worldline:live when dryRun: false and the request supplies approving attestationIds over the persisted comparison-artifact:* baseline. Successful execution still lowers through the shared mutation kernel rather than a special collapse engine, including committed content-clearing transfer ops, and returns a terminal live observation plus mutation execution details.

• collapse_worldline runway over shared transfer planning — xyph api collapse_worldline now maps onto git-warp’s published coordinate transfer planning surface for canonical derived worldlines. The command requires a fresh comparisonArtifactDigest, recomputes the source-vs-target comparison to detect drift, supports settlement against worldline:live, lowers the resulting transfer plan through the shared mutation kernel, and returns a typed XYPH collapse-proposal with per-side observations, sanitized transfer ops, substrate digests, and either mutation preview or mutation execution details.

• Durable collapse-proposal:* governance records — xyph api collapse_worldline now accepts persist: true to record the current collapse artifact on worldline:live. The durable node stores indexed proposal metadata plus a deterministic JSON body containing the preview or execution payload and published git-warp fact exports, making the proposal a durable governance record without inventing new substrate serialization

• Durable comparison-artifact:* governance records — xyph api compare_worldlines now accepts persist: true and records a deterministic comparison artifact on worldline:live, carrying both the XYPH operationally scoped git-warp fact and the raw whole-graph git-warp fact in the recorded body.

• Thin braid_worldlines mapping over published substrate braids — xyph api braid_worldlines now targets a canonical derived worldline, pins one or more canonical derived supportWorldlineIds through git-warp’s published braid substrate, accepts optional readOnly, and returns XYPH-first braid/worldline metadata plus substrate backing identifiers. It changes co-present visibility without pretending merge, rebase, or collapse semantics

• Operationally scoped compare_worldlines preview — xyph api compare_worldlines now maps onto git-warp's published coordinate comparison surface twice: once for raw whole-graph audit truth and once for XYPH's operationally scoped freshness truth. The command returns a typed XYPH comparison-artifact preview with deterministic artifact identity, per-side observation coordinates, substrate-backed visible patch divergence, visible node/edge/property deltas, optional target-local comparison detail when targetId is provided, and durable recording support without self-perturbing tip-vs-tip freshness.

• Working-set-backed derived-worldline execution — canonical derived worldlines now support honest working-set-local observe(graph.summary), observe(worldline.summary), observe(entity.detail), history, diff, and apply. These reads materialize the backing git-warp working set rather than the shared live graph, and apply lowers through the same mutation kernel as live writes while committing to the working-set overlay patch log. Observation coordinates across these commands now reflect the working-set-visible frontier instead of quietly reporting the live graph digest

• Braided derived-worldline parity across the canonical control-plane slice — braided canonical derived worldlines now stay internally consistent across observe(graph.summary), observe(worldline.summary), observe(entity.detail), history, diff, apply, and observe(conflicts), with integration coverage proving the selected braid yields one coherent execution surface.

• Working-set-backed fork_worldline — xyph api fork_worldline now creates derived XYPH worldlines by delegating to git-warp working-set creation. The current slice is intentionally narrow: it supports worldline:live as the source worldline, accepts optional at: { tick } as a current-frontier Lamport ceiling, and returns an XYPH worldline payload plus the backing git-warp working-set identifier without pretending arbitrary historical frontier or nested derived-worldline support already exists

• Worldline-aware substrate conflict projection — xyph api observe now supports projection: "conflicts" backed by git-warp's read-only analyzeConflicts() surface for the live frontier or a canonical derived worldline's backing working set. The projection remains tip-only, accepts optional lamportCeiling and analyzer filters, and returns substrate conflict traces without inventing XYPH-side counterfactual truth

• Historical observation selectors and structured redaction — xyph api low-level reads now support at / since tick selectors for observe(graph.summary), observe(worldline.summary), observe(entity.detail), history, and diff, using isolated read graphs materialized at explicit ceiling ticks. Sealed observation now returns structured redactions metadata for content-bearing entity.detail payloads instead of treating partial withholding as total failure

• Per-request capability resolution for xyph api — the sovereign control-plane now computes principal, observer profile, and effective capability grants per request instead of assuming one process-wide identity. apply, attest, propose, comment, and hidden admin/debug commands now flow through a shared capability resolver, and audit/observation metadata carry the resolved principal and capability context

• Sovereign control-plane foundation — added xyph api, a versioned JSONL machine interface backed by ControlPlaneService with observe, history, diff, explain, fork_worldline, compare_worldlines, collapse_worldline, apply, comment, propose, and attest command support. query and rewind_worldline remain reserved without pretending they are implemented yet

• Generic mutation kernel — added MutationKernelService, a single allowlisted primitive-op mutation path (add_node, remove_node, set_node_property, add_edge, remove_edge, set_edge_property, attach_node_content, clear_node_content, attach_edge_content, clear_edge_content) for sovereign control-plane writes and future collapse lowering

• Append-only proposal and attestation records — added RecordService support for proposal:* and attestation:* durable records, alongside comment creation routed through the same mutation kernel instead of ad hoc CLI patch logic

• Canonical artifact record helper — RecordService now supports durable canonical control-plane artifact recording for XYPH-owned node families such as collapse-proposal:*, keeping ontology/governance meaning in XYPH while reusing published git-warp fact exports in the recorded body

• Canonical sovereign-control-plane docs — public docs now state that XYPH's ontology is sovereign, that observer profiles do not grant authority by existing, and that Alfred-derived components may sit behind ports without leaking into XYPH's public API vocabulary

• xyph doctor graph health audit — new CLI command audits dangling edges (including incoming edges from missing nodes), orphaned workflow/narrative/traceability nodes, readiness contract gaps, sovereignty violations, and governed completion gaps. Supports both human-readable output and --json for automation

• xyph doctor prescribe deterministic remediation view — doctor now derives structured prescriptions with category (structural-blocker, structural-defect, workflow-gap, hygiene-gap), blocked transitions, blocked task IDs, effective priority inheritance, and top remediation buckets for automation and agent-facing triage

• Doctor-backed recommendation work in the agent protocol — briefing, next, and context now surface doctor prescriptions as recommendation work instead of burying graph-health findings inside diagnostics only. briefing includes a recommendationQueue, context includes recommendationRequests, and next can surface urgent doctor-driven inspect candidates

• Global "My Stuff" drawer — press m from any screen to toggle an animated drawer showing agent's quests, submissions, and recent activity. Slides in from the right with a tween animation; content is agent-scoped when XYPH_AGENT_ID is set. Replaces the fixed right column on the dashboard view

• Campaign DAG visualization — campaigns with inter-campaign dependencies are rendered as a mini-DAG using bijou's dagLayout(), sorted topologically. Falls back to flat list when no dependencies exist

• Status bar progress bar — compact gradient progress bar added to the status line showing quest completion percentage

• Graph algorithm ban enforcement — scripts/check-graph-algorithms.sh detects userland BFS/DFS/topo-sort/Dijkstra patterns in src/. Wired into CI (strict-policy job) and pre-commit hook. Enforces the rule that all graph traversals must use git-warp's graph.traverse.* API

• transitiveDownstream in GraphSnapshot — pre-computed via graph.traverse.bfs() in GraphContext; eliminates the last userland BFS in computeTopBlockers() (DepAnalysis.ts)

Fixed

• PR feedback hardening across summaries, quest detail, agent case work, and suggestion actions — control-plane summary counts now include milestones in the campaign census, analysis-profile snapshots retain governed case links for AI suggestions, milestone-backed quest detail reads preserve their parent context, missing criterion verifiable flags now default to true consistently in quest detail, narrative/comment expansion now uses graph.traverse.bfs() instead of hand-rolled queue walkers, next() now keeps an inspect fallback for cases that are ready for judgment but not briefable, and the TUI blocks empty adopt/supersede suggestion rationales before dispatching writes.
• PR feedback hardening across derived reads, agent queues, and suggestion resolution — per-profile snapshot caches no longer let stale full snapshots survive newer operational reads, derived worldline content detail no longer leaks mismatched live-tip bodies when the derived _content OID diverges, agent briefing case queues now reuse the same operational read boundary as the briefing snapshot, Ask-AI from a governed case page now binds to the open case instead of the underlying lane selection, and both CLI/TUI suggestion adoption or supersession now require an explicit non-empty rationale. The lockfile was also refreshed to clear the current production npm audit findings.
• Derived worldline graph shims now use an explicit GraphContext read surface — createGraphContextFromGraph() now accepts the smaller read-only surface it actually consumes, GraphContext guards syncCoverage() explicitly instead of assuming every caller is a full live graph handle, and the derived control-plane adapter no longer suppresses the type system with as unknown as WarpGraph just to satisfy entity-detail reads.
• Inspector scroll status and confirm modal hint rendering were clarified — the inspector no longer shows BIJOU’s ambiguous raw Line N/M footer label. XYPH now renders a clearer Scroll N/M status only when the inspector detail actually overflows, drives the inspector rail from wrapped pager line counts, and explicitly terminates styled confirm-hint segments so the quit modal border no longer inherits the final red cancel color.
• act merge now fails loudly when decision recording does not land — if the git merge succeeds but XYPH cannot write the authoritative merge decision back into the graph, the agent action kernel now returns a non-success partial-failure outcome and the CLI exits with an error envelope instead of reporting the merge as fully successful. This keeps automation retryable and prevents silently “settled” submissions that never recorded their decision state
• Trust test fixture is now tracked correctly — .gitignore now ignores only the repo-root trust/ and .xyph/ directories, so test/fixtures/trust/keyring.json is committed and available in CI. Fixes patch-ops matrix failures caused by missing public-key fixture data on clean checkouts
• Blocker counts exclude GRAVEYARD tasks — transitiveDownstream BFS now excludes GRAVEYARD tasks from counts (matching DONE exclusion), and filterSnapshot() strips GRAVEYARD keys from the map. Fixes inconsistency where directCount (from filtered edges) and transitiveCount (from unfiltered BFS) could diverge when --include-graveyard was not set
• Landing screen auto-dismiss — showLanding is now set to false when snapshot-loaded fires, so the dashboard appears immediately after data loads. Previously required a throwaway keypress to dismiss the landing screen, making navigation appear broken
• View cycling via [/] — bracket keys cycle forward/backward through all 5 views from anywhere, matching bijou's createFramedApp convention ([/] = tabs, Tab = panes). Tab/Shift+Tab retained for panel switching within views
• Transactional updateKeyring API (KeyringStoragePort) — new updateKeyring(mutator) method with KeyOps handle; adapter records an undo log and rolls back private-key side-effects in reverse order on failure (KSP-001)

Changed

• --json now speaks JSONL stream semantics — CLI JSON mode can emit newline-delimited JSON event records before the terminal success/error envelope. xyph doctor and xyph doctor prescribe now stream start and progress records while long audits run; the final result record remains the existing success/error envelope shape

• Agent action validation now refuses structurally illegal transitions — the action kernel consults cached doctor prescriptions before ready, submit, review, seal, and merge; transitions blocked by impossible or broken graph state now fail with illegal-graph-state instead of limping forward

• Dashboard restyled to single-column layout — graph stats use bold primary labels instead of dim/muted text; health metrics (sovereignty, orphans, forked) merged into the stats area; "Inbox" label renamed to "Backlog"; Top Blockers rendered as bijouTable() instead of enumeratedList()

• Dashboard panel switching removed — Tab/Shift+Tab on dashboard is now a no-op. The right column content moved to the global drawer (m key)

• Upgrade bijou 1.3.0 → 1.8.0 — all three packages (@flyingrobots/bijou, bijou-node, bijou-tui) bumped. Test helper imports updated to use @flyingrobots/bijou/adapters/test subpath

• Upgrade git-warp 13.1.0 → 14.0.0 — major version bump, no breaking changes for current usage

• Upgrade git-warp 14.0.0 → 14.2.0 — XYPH now consumes the published substrate conflict analyzer and raises its explicit dependency floor accordingly

• Upgrade git-warp 14.2.0 → 14.3.0 — XYPH now tracks the published substrate release that expands the git-warp time-travel debug CLI and its debugger architecture/docs, keeping the dependency floor aligned with the released substrate toolchain

• Upgrade git-warp 14.3.0 → 14.4.0 — XYPH now tracks the published substrate release that expands the built-in git warp debug time-travel CLI with coordinate and timeline inspection, keeping the released debugger surface aligned with the substrate version XYPH depends on

• Upgrade git-warp 14.4.0 → 14.5.0 — XYPH now tracks the published substrate release that adds the working-set foundation: explicit coordinate materialization plus durable git warp working-set management for pinned substrate coordinates, without forcing XYPH to consume unreleased git state

• GuildSealService simplified — generateKeypair() and rotateKey() now delegate rollback to updateKeyring(), removing hand-written try/catch rollback choreography from the domain layer

• Shared test helpers (test/helpers/) — makeSnapshot(), entity builders (quest, intent, campaign, scroll, submission, review, decision), strip() ANSI helper, makeKey()/makeResize() keyboard factories, and mock port factories (mockGraphContext, mockIntakePort, mockGraphPort, mockSubmissionPort). Eliminates duplication across TUI test files

• Shared view helpers (src/tui/view-helpers.ts) — sliceDate(), groupBy() eliminate repeated date formatting and array grouping patterns across render-status.ts and bijou view files

• assertNodeExists() validator (src/cli/validators.ts) — shared graph node existence check replacing 15 inline hasNode() + throw NOT_FOUND patterns across 5 command files

• SecretPort interface (src/ports/SecretPort.ts) — extracted SecretAdapter interface from infrastructure to ports layer, fixing hexagonal architecture boundary violation (H3)

• KeyringStoragePort — new port abstracting keyring and private-key persistence (src/ports/KeyringStoragePort.ts), decoupling GuildSealService from node:fs, node:path, and node:crypto

• FsKeyringAdapter — filesystem-backed implementation of KeyringStoragePort (src/infrastructure/adapters/FsKeyringAdapter.ts)

• TestParserPort — new port abstracting test-file parsing (src/ports/TestParserPort.ts), decoupling analysis pipeline from the TypeScript Compiler API

• TsCompilerTestParserAdapter — TS Compiler API implementation of TestParserPort (src/infrastructure/adapters/TsCompilerTestParserAdapter.ts), containing logic previously in src/domain/services/analysis/TestFileParser.ts

• scripts/ under lint + typecheck — tsconfig.scripts.json and ESLint config now cover all TypeScript scripts; npm run lint checks scripts alongside src/test

• Consolidated wire-deps.ts — single idempotent dependency-wiring script replacing wave2/wave3/fixup; gracefully skips missing nodes, detects cycles and duplicates

• migrate-voc-001.ts — one-shot graph migration script, patched 116 legacy INBOX nodes to BACKLOG

Changed

• RoadmapPort split — RoadmapQueryPort (reads), RoadmapMutationPort (writes), RoadmapSyncPort (sync) with RoadmapPort as their union for backward compat (S6/ISP)
• GuildSealService refactored to port-based architecture — all filesystem and crypto RNG operations extracted to KeyringStoragePort; domain service is now pure, depending only on the port interface (hex audit H1 fix)
• TestFileParser moved to infrastructure layer — implementation relocated from src/domain/services/analysis/ to TsCompilerTestParserAdapter; domain file replaced with deprecation tombstone (hex audit H2 fix)
• Keybindings standardized to bijou v1.6.0 vim conventions — all views now use d/u for half-page scroll, g/G for jump-to-top/bottom, space toggles accordion in lineage view. Reject in backlog rebound from d to D (shift+d) to avoid conflict with page-down. Command palette navigation changed from j/k to Ctrl+N/Down and Ctrl+P/Up (j/k now type filter characters). All view keymaps built on bijou preconfigured factories (navTableKeyMap, accordionKeyMap, commandPaletteKeyMap)
• Upgrade bijou v1.3.0 → v1.6.0 — all three packages (@flyingrobots/bijou, @flyingrobots/bijou-node, @flyingrobots/bijou-tui) bumped to v1.6.0. New capabilities include: interactiveAccordion(), transition shaders (7 built-in page transitions), preconfigured *KeyMap() factories for all building blocks, markdown() terminal renderer, hyperlink() (OSC 8), dagStats(), auditStyle test adapter, drawer with all 4 anchors, and DTCG theme interop
• VOC-001: Vocabulary rename — raw graph statuses now match the domain model. inbox command writes BACKLOG (was INBOX), promote writes PLANNED (was BACKLOG), quest/quest-wizard write PLANNED (was BACKLOG), reopen writes BACKLOG (was INBOX). normalizeQuestStatus() retained as legacy shim for un-migrated nodes only
• Reject from BACKLOG or PLANNED — reject command now accepts quests in BACKLOG or PLANNED status; adapter and domain service aligned
• Scripts migrated to git-warp v13 API — all props.get('key') calls in scripts/ replaced with Record bracket notation props['key']; Array<T> → T[], unused imports removed
• Quest status lifecycle diagram — updated to reflect VOC-001 rename (no more INBOX state)

Fixed

• WarpIntakeAdapter promote/reject guards — adapter and domain service now use the same status vocabulary; promote checks BACKLOG, reject accepts BACKLOG|PLANNED

Removed

• VALID_RAW_STATUSES — no longer needed; normalizeQuestStatus() handles legacy values
• task:BX-017 (HistoryPort) — graveyarded as redundant; git-warp v13 natively provides patchesFor(), materializeSlice({ receipts: true }), and materialize({ ceiling }). Downstream quests (BX-009–016) call git-warp directly
• 14 obsolete/duplicate quests graveyarded — VOC-002 (normalization already done), cli-backlog-add (duplicate of inbox), upstream-ink-fullscreen & bijou-v09-title-refactor & bijou-type-guards & e2e-dashboard-smoke (Ink removed), BKL-PRESET-SOT & inline-color-status (already satisfied), suggestion-calibrate & IDEA-TEMPORAL-TRACE & temporal-traceability-queries & IDEA-HEATMAP & traceability-heat-map & IDEA-SCAN-IMPL (duplicates)

Previously Added

• Key rotation — GuildSealService.rotateKey(agentId) generates a new Ed25519 keypair, marks the previous key as retired (active: false), and registers the new key as the sole active key for the agent. Retired keys stay in the keyring for verification of historical signatures (WVR-006)
• Pre-rendered SVG diagrams — 24 diagrams across 22 documentation files, rendered from Mermaid source (.mmd) to SVG via mmdc. Source files in docs/diagrams/*.mmd, rendered output in docs/diagrams/*.svg. Render script: scripts/render-diagrams.sh. Covers entity relationships, state machines, flowcharts, sequence diagrams, and architecture stacks (WVR-006)
• history command — xyph-actuator history <nodeId> shows all patches that touched a node via git-warp's patchesFor() provenance API (Constitution Art. III compliance)
• Multibase DID key encoding — encodeBase58btc() and publicKeyToDidKey() in crypto.ts; Ed25519 public keys are now encoded as spec-compliant did:key:z6Mk... identifiers using the multicodec prefix 0xed01 + base58btc multibase encoding (WVR-006)
• Versioned keyring migration pipeline — loadKeyring() transparently migrates older keyring formats on read via a sequential migration chain (v1 → v2 → ...); future schema changes (key rotation, expiry, multi-algorithm) slot in by appending a migration function (WVR-006)
• Guild Seals documentation — docs/GUILD_SEALS.md covers the full signing system top-to-bottom: key generation, DID key encoding, signing flow, verification, keyring schema, versioning, migration pipeline, trust directory layout, security considerations, and FAQ (WVR-006)

Changed

• Keyring schema v3 — KeyringEntry gains active boolean for key rotation support; loadKeyring() validates at-most-one active key per agent (multiple retired keys allowed); v2→v3 migration defaults all existing entries to active: true (WVR-006)
• Keyring schema v2 — KeyringEntry gains agentId (agent→key lookup) and legacyKeyIds (alias resolution for backward-compatible verification); loadKeyring() indexes the Map by both canonical derived keyId and legacy aliases so old patches and new seals both resolve correctly (WVR-006)
• keyIdForAgent() derives real DID keys — GuildSealService.keyIdForAgent() now looks up the agent's public key from the keyring and derives a proper multibase-encoded did:key instead of the previous did:key:<agentId> placeholder (WVR-006)
• sign() derives keyId from private key — GuildSealService.sign() now computes the public key from the private key at sign time and derives the DID key directly, ensuring the seal's keyId always matches the signing key
• generateKeypair() writes v2 keyrings — new keypairs are written with the derived did:key:z6Mk... as the canonical keyId and include the agentId field (WVR-006)
• StylePort interface — dependency-injected styling abstraction (StylePort) with BijouStyleAdapter (chalk-based) and PlainStyleAdapter (no-color) implementations; replaces the global theme singleton with explicit wiring through the composition root
• Theme lab TUI — interactive design token exploration tool (xyph-theme-lab.ts) for cycling through theme palettes in real time
• Dark + light theme variants — automatic terminal background detection with distinct dark and light palettes; theme selection adapts to the user's terminal

Changed — StylePort Migration

• CLI pipeline uses StylePort DI — all CLI commands and TUI views receive StylePort via the composition root instead of importing a global theme module; threading flows from xyph-actuator.ts → commands → render functions

Removed

• Ink / React / JSX — removed ink, react, @types/react, boxen, and cli-table3 dependencies (all replaced by bijou). Removed JSX compiler options (jsx, jsxImportSource) from tsconfig. Renamed xyph-dashboard.tsx → xyph-dashboard.ts. No JSX or React code remains in the codebase.
• Global theme bridge — deleted src/tui/theme/bridge.ts and its tests. All styling now flows through the dependency-injected StylePort interface.
• Dead theme barrel module — deleted orphaned src/tui/theme/index.ts barrel re-export

Fixed

• Schema keyId pattern sync — schemas/PATCH_OPS_SCHEMA.v1.json keyId regex now matches the multibase Base58btc spec (z[1-9A-HJ-NP-Za-km-z]{10,100}), replacing the loose z6[A-Za-z0-9]+ that accepted invalid multibase characters (WVR-006)
• Schema deduplication — docs/canonical/PATCH_OPS_SCHEMA.json replaced with a symlink to schemas/PATCH_OPS_SCHEMA.v1.json; the runtime schema is now the single source of truth, eliminating a persistent drift trap (WVR-006)
• Schema documentation — added description to baseSnapshotDigest in the schema (advisory, not a concurrency gate); added deprecated + description to rollbackPatchDigest in AUDIT_EVENT_SCHEMA.json (WVR-006)
• Prose alignment with CRDT reality — replaced centralized-database vocabulary in README (atomicity/rollback), WHITEPAPER ("work is a transaction"), EXECUTIVE_SUMMARY (signed → attributed, 900→650 tests), AGENT_CHARTER (blockedBy → depends-on), ORCHESTRATION_SPEC (snapshot mismatch → entity conflict), PATCH_OPS_INVARIANTS (baseSnapshotDigest advisory note) (WVR-006)
• DashboardApp watching lifecycle — startWatching() (graph.watch polling) now fires from init() alongside the initial fetchSnapshot(), instead of being conditionally triggered on the first snapshot-loaded message; the watching model field still tracks state but no longer gates command emission
• Test fixture type completeness — added missing watching: false to the makeModel() helper in views.test.ts to match the updated DashboardModel type
• graph.watch() process hang — stopWatching() now clears the poll interval on quit, preventing the Node process from hanging after dashboard exit
• Theme lab bridge bypass — theme lab creates its own StylePort instances directly instead of routing through the (now-deleted) global bridge
• Stale .tsx doc comment — fixed leftover .tsx reference in xyph-dashboard.ts
• Coordinator daemon StylePort hoisting — StylePort instance lifted to module level in coordinator-daemon.ts to avoid repeated construction

Fixed — PR #36 Code Review

• Lineage view hid orphan quests — renderLineage returned early when snapshot.intents.length === 0, skipping orphan quest rendering; orphan detection now runs before the early return so sovereignty violations are always visible
• Remote-change events dropped during fetch — DashboardApp now sets refreshPending: true when a remote-change arrives while a snapshot fetch is in-flight, triggering a follow-up refresh on completion
• Theme resolver read env var twice — BijouStyleAdapter created a second createThemeResolver that re-read XYPH_THEME; now uses the already-resolved theme as fallback without re-reading
• Gradient sample RangeError — xyph-theme-lab.ts could pass negative value to '#'.repeat() on narrow terminals; clamped to Math.max(0, ...)
• Double-styling in history command — removed chalk.dim() inside ctx.muted() in coordination.ts; styling now flows exclusively through StylePort
• Guild terminology in roadmap — replaced "tasks" with "Quests" in frontier blocked/waiting messages
• ARCHITECTURE.md diagram — updated (TUI/Ink) label to (TUI/TEA) to match the bijou migration
• Lint scope — added xyph-theme-lab.ts to ESLint config, lint script, and tsconfig.json includes; fixed unused w parameters surfaced by strict compilation
• Stale bridge.ts references — updated comments in xyph-presets.ts to reference BijouStyleAdapter
• Theme brightness test — replaced packed-RGB comparison with W3C relative luminance formula
• Test header assertion — removed toContain('Backlog') header text assertion per project test guidelines

Added — Work DAG Analysis Suite

• DagAnalysis.ts — pure functions for DAG structure analysis: DAG width, greedy worker scheduling, and anti-chain grouping
• scripts/generate-work-dag.ts — generates comprehensive DAG visualization suite: full/per-campaign/backlog/graveyard SVGs in both LR and TB orientations, plus work.md analysis document with topological sort, critical path, parallelism, scheduling, transitive reduction/closure, ancestry/impact, campaign grouping, and anti-chain waves
• npm run graph:work — runs the generator, outputs to docs/work/
• 17 unit tests — unit tests for remaining DagAnalysis functions (diamond, linear, empty, single-node, isolated-node graphs)

Changed

• Upgraded git-warp to v13.1.0 — @git-stunts/git-warp from v12.1.0 to v13.1.0; migrated all getNodeProps() call sites from Map<string, unknown> (.get()) to Record<string, unknown> (bracket notation)
• Delegated graph algorithms to git-warp traversals — removed userland computeLevels(), transitiveReduction(), transitiveClosure(), reverseReachability(), computeProvenance() from DagAnalysis.ts; generate-work-dag.ts now calls graph.traverse.levels(), .transitiveReduction(), .transitiveClosure(), .rootAncestors() directly; computeAntiChains() simplified to accept pre-computed levels
• Upgraded bijou to v1.3.0 — @flyingrobots/bijou, @flyingrobots/bijou-node, @flyingrobots/bijou-tui from v1.2.0 to v1.3.0
• Upgraded bijou to v1.2.0 — @flyingrobots/bijou, @flyingrobots/bijou-node, @flyingrobots/bijou-tui from v0.10.0 to v1.2.0
• Roadmap DAG panel uses dagPane() — replaced ~50 lines of manual dagLayout() + viewport() + scroll-centering math with bijou v1.2.0's dagPane() building block; auto-scroll-to-selection, keyboard-synced DAG highlight, DagPaneState replaces raw dagScrollX/dagScrollY
• Dashboard columns use focusArea() — replaced bare viewport() with focusArea() for visual focus indication; focused column shows bright ▎ gutter, unfocused shows muted gutter
• Default theme — switched from cyan-magenta to teal-orange-pink

Fixed — PR #34 Feedback

• Vitest config simplification — removed explicit include globs from vitest.config.ts and retained focused exclude rules, using Vitest default test discovery for future *.spec.* / *.test.tsx compatibility
• CI audit vulnerability (tar) — updated lockfile resolution from tar@7.5.8 to tar@7.5.10 (via transitive @mapbox/node-pre-gyp/node-gyp) so npm audit --omit=dev passes with zero vulnerabilities

Fixed — PR #33 Code Review

• Roadmap fallback paging restored — PageDown/PageUp now scroll the right roadmap panel even when a snapshot has no dependency edges; fallback viewport scroll is stateful instead of hardcoded to row 0
• DagPane always built when quests exist — snapshot refresh now creates roadmap.dagPane for quest-only graphs, while critical-path computation remains gated on dagEdges.length > 0
• Defensive DagPane sizing — DagPane width/height are clamped to positive minimums on both resize and snapshot rebuild paths
• Unused analysis arg removed — dropped the unused sorted parameter from buildAnalysisInputs() in scripts/generate-work-dag.ts and updated call sites
• SVG ignore intent documented — .gitignore targets generated DAG artifacts (docs/assets/work-dag*.svg) and now includes inline rationale for the pattern

Fixed — PR #32 Code Review

• DONE tasks inflated scheduling makespan — scheduleWorkers now treats DONE tasks as weight 0, matching computeCriticalPath semantics (Codex P1)
• CI traceability job failure — added fetch-depth: 0 and git identity config to traceability workflow; shallow clones lack commit objects needed by git-warp materialization
• Topological sort delegated to graph traversal API — generate-work-dag.ts now delegates to graph.traverse.topologicalSort() instead of using a hand-rolled Kahn implementation (P1-01)
• Unused reverseReachability import — removed from generate-work-dag.ts (P2-01)
• Dead allNodes set in computeProvenance — removed unused variable construction loop (P2-02)
• Non-null assertions — replaced ! assertions in generate-work-dag.ts and DagAnalysis.test.ts with guard patterns (P3-01/02/04)
• Redundant STATUS_COLORS['BACKLOG']! — extracted DEFAULT_COLORS constant, eliminating non-null assertion (P3-03)
• 237KB SVG tracked in repo — removed docs/assets/work-dag.svg from index and added a generated-artifact ignore rule (docs/assets/work-dag*.svg) to .gitignore (P4-01)

Added — Workflow Infrastructure

• Git hooks — scripts/hooks/pre-commit (lint gate) and scripts/hooks/pre-push (test gate); plain shell scripts, no Husky/lint-staged
• npm run graph:pull — fetches WARP writer refs from origin for local materialization
• npm run graph:push — pushes WARP writer refs (refs/warp/xyph-roadmap/writers/*) to origin for CI and collaborator access; excludes checkpoint/coverage cache refs which are rebuilt locally
• CI traceability job — new traceability job in .github/workflows/ci.yml fetches WARP refs and runs analyze --dry-run --json for coverage reporting
• .xyph.json in .gitignore — local config file excluded from version control
• Self-referential analyze baseline — ran analyze --dry-run against XYPH's own 745-test suite; 49 files scanned, 0 graph targets (traceability chain not yet populated)
• Backlog items — 10 new inbox tasks: soft-gate merge, TUI suggestion tab, dashboard suggestion widget, roadmap coverage badge, traceability heat map, production code scan annotations, temporal traceability queries, auto graph push hook, CI graph cache, suggestion calibration
• M11 Phase 3 design doc — docs/M11-phase3-design.md covers computed status propagation, DoD policies, hard-gate seal/merge, and gap detection

Fixed — PR #29 Code Review

• Evidence dedup uses sourceFile — EvidenceNode now exposes sourceFile from graph props; analyze filters existing evidence by sourceFile:criterionId instead of producedBy:criterionId (C-1)
• Consolidated LayerScore interface — single canonical definition in analysis/types.ts, re-exported from Suggestion.ts; eliminates duplicate interface (M-1)
• Batch graph patches in analyze — auto-link and suggestion writes use one graph.patch() call each instead of per-item patches (M-2)
• Batch graph patches in suggestion accept-all — single graph.patch() for all accepted suggestions (M-3)
• Removed double casts in ConfigAdapter — exported mergeWeights/mergeLlm from ConfigResolution, used to safely parse partial JSON from graph (m-2)
• Type-safe config set — replaced as never cast with exhaustive switch on config key (m-3)
• Inlined global regex in TraceabilityScan — CRITERION_REF regex now local to scanAnnotations(), avoids stale lastIndex from global flag (m-4)
• Removed duplicate 'should' in STOP_WORDS — ImportDescribeLayer had 'should' listed twice (N-2)
• Fixed renderAll total count — now includes submissions, reviews, decisions, stories, requirements, criteria, evidence, and suggestions (N-3)
• Hardened Suggestion constructor — per-entry LayerScore validation (layer, score, evidence shape checks) and deep-freeze of individual layer objects
• Hardened GraphContext suggestion parsing — Number.isFinite() guard on confidence, per-element shape validation on parsed layer JSON
• Rendered rejected suggestions — renderSuggestions now shows a Rejected section with rationale, parallel to Accepted
• Campaign cardinality in orphan script — assign-orphan-campaigns.ts now validates campaign existence and skips quests already assigned to another campaign instead of adding duplicate belongs-to edges
• Hash-based evidence/suggestion IDs — analyze now uses SHA-256 hash of testFile+targetId instead of slug truncation, preventing ID collisions across files
• Removed unsafe casts in config CLI — config get/set/list output no longer uses as unknown as double casts
• Criterion validation in scan — scan command now validates criterion node existence before writing evidence edges, skips missing criteria with a warning, and batches writes into a single patch
• TestFileParser handles modifiers — getCallName now resolves it.only, describe.skip, it.each(...)() to their root identifier
• LLM adapter graceful fallback — AnthropicLlmAdapter.getSecret() moved inside try/catch so vault errors trigger graceful degradation instead of crashing
• Evidence dedup for implements edges — EvidenceNode now exposes requirementId from implements edges; analyze dedup filter covers both verifies (criterion) and implements (requirement) edges
• Synced package-lock.json — lock file now includes @anthropic-ai/sdk, @git-stunts/vault, and transitive deps; fixes all CI npm ci failures

Added — M11 Phase 4: Intelligent Test Auto-Linking

• Config infrastructure — layered resolution (env > .xyph.json > graph > defaults), config get/set/list CLI commands (ALK-001)
• suggestion: node type — auto-detected test→criterion/requirement links with PENDING/ACCEPTED/REJECTED lifecycle, LayerScore breakdown (ALK-002)
• Suggestion lifecycle commands — accept, reject, accept-all CLI commands with evidence materialization on accept (ALK-003)
• Test file parser — TypeScript Compiler API extraction of imports, describe/it blocks, function/method calls (ALK-004)
• Heuristic scoring framework — ScoreCombiner with weighted averaging and automatic renormalization for missing layers (ALK-005)
• 4 heuristic layers — FileNameLayer (0.4–0.8), ImportDescribeLayer (0.3–0.7), AstLayer (0.7–0.9), SemanticLayer (Jaccard ×3) (ALK-006)
• LLM port + Vault integration — provider-agnostic LlmPort, VaultSecretAdapter for OS keychain, AnthropicLlmAdapter (ALK-007)
• LLM heuristic layer — batch test-target analysis with structured JSON matching, graceful fallback (ALK-008)
• xyph analyze command — full orchestration: config→glob→parse→score→classify→write, with --dry-run, --layers, --min-confidence (ALK-009)
• status --view suggestions — pending/accepted/rejected suggestions with confidence and layer breakdown (ALK-010)
• suggests edge type — suggestion→target linking in graph schema
• 107 new tests — 10 test suites covering all new domain services, entities, and analysis layers
• Dependencies — @git-stunts/vault 1.0.1, @anthropic-ai/sdk 0.78.0

Added — M11 Traceability Phases 1 & 2

• story: node type — user stories with persona/goal/benefit, story actuator command (TRC-001)
• req: node type — requirements with kind (functional/non-functional) and MoSCoW priority, requirement actuator command (TRC-001)
• criterion: node type — acceptance criteria attached to requirements, criterion actuator command (TRC-005)
• evidence: node type — verification evidence (test/benchmark/manual/screenshot) with pass/fail result, evidence actuator command (TRC-005)
• decomposes-to edge type — intent→story and story→req decomposition, decompose actuator command with cycle checks (TRC-002)
• has-criterion edge type — requirement→criterion attachment (TRC-005)
• verifies edge type — evidence→criterion verification link (TRC-005)
• GraphSnapshot extended — stories[], requirements[], criteria[], evidence[] fields with full edge resolution (TRC-003, TRC-006)
• status --view trace — traceability chain renderer: stories→reqs→criteria→evidence with coverage stats (TRC-004)
• xyph scan — test annotation parser (// @xyph criterion:<id>) that writes evidence nodes automatically (TRC-007)
• Completeness queries — computeUnmetRequirements, computeUntestedCriteria, computeCoverageRatio pure functions (TRC-008)
• 57 new tests — 4 entity unit tests, scan parser, analysis queries, full integration round-trip

Added — M11 Traceability Decomposition

• intent:TRACEABILITY — sovereign intent for requirements, acceptance criteria, and evidence as first-class graph objects
• 13 TRC quests (TRC-001..013) under campaign:TRACE, spanning 4 phases: Foundation, Criteria & Evidence, Computed Status, Intelligence
• 12 dependency edges wiring the TRC quest execution order
• Rejected task:traceability-m11 placeholder — superseded by the concrete decomposition

Added — Orphan Campaign Assignments

• 58 orphan quests assigned to campaigns via scripts/assign-orphan-campaigns.ts:
  • 22 → campaign:DASHBOARD (TUI, bijou, rendering)
  • 24 → campaign:CLITOOL (CLI, CI, testing, theme)
  • 6 → campaign:ECOSYSTEM (external integrations, scaling)
  • 4 → campaign:AGENT (agent protocol)
  • 1 → campaign:SUBMISSION, 1 → campaign:BEDROCK
• Zero orphan quests remaining (was 30% of all quests)

Fixed

• Cross-type depend guard — depend now rejects edges between different type families (e.g. task: → campaign:); both nodes must be tasks, or both must be campaigns/milestones
• CHANGELOG version ordering — restored [Unreleased] → [alpha.12] → [alpha.11] sequence
• --json deps output — milestoneExecutionOrder now included in status --view deps --json output
• Milestone frontier with zero dep edges — computeFrontier is now always called for campaigns; previously returned empty frontier when no depends-on edges existed, hiding actionable milestones

Changed

• Replaced unsafe as cast in seed-milestone-deps.ts with toNeighborEntries() runtime validator
• Removed redundant optional-chaining in render-status.ts milestone sections (extracted to local const for proper narrowing)
• Array<[string, string]> → [string, string][] style fix in seed script

[1.0.0-alpha.12] — 2026-03-03

Added — Milestone Dependencies

Campaign dependency graph:

• CampaignNode now carries description and dependsOn fields
• GraphSnapshot includes sortedCampaignIds (topological order via git-warp traversal)
• depend command widened to accept campaign: and milestone: nodes alongside task: nodes
• Cycle detection works across all node types

Milestone frontier in deps view:

• status --view deps now shows Milestone Frontier (campaigns whose deps are all DONE) and Milestones Blocked tables
• --json output includes milestoneFrontier, milestonesBlocked, and milestones fields
• Dashboard campaigns section annotates blocked campaigns with [blocked] indicator

Data migration:

• Seeded descriptions on all 13 campaigns, created missing campaign:TRACE (M11) and campaign:ECOSYSTEM nodes
• Added 8 inter-milestone depends-on edges
• Fixed campaign:WEAVER status from BACKLOG to DONE
• Removed docs/ROADMAP.md — milestone data now lives in the WARP graph

[1.0.0-alpha.11] — 2026-03-02

Changed

Navigation model rework:

• Number keys 1–5 now jump directly to views (dashboard/roadmap/submissions/lineage/backlog), replacing Tab/Shift+Tab view cycling
• Tab is rebound on the dashboard view to switch focus between left (In Progress) and right (My Quests) panels
• PageDown/PageUp on the dashboard now scroll the focused column instead of toggling panels
• Command palette shortcuts updated to show 1–5 for view switching

Independent column scrolling:

• Dashboard left and right columns are now wrapped in viewport() from bijou-tui, enabling vertical scrolling when content overflows the terminal height
• New leftScrollY/rightScrollY state tracks scroll position per column independently
• Health, Top Blockers, and Activity Feed sections are now reachable via PageDown on tall dashboards

Confirm overlay:

• confirmOverlay() now accepts an optional custom hint string (used by quit confirmation dialog)

Quit flow redesign:

• Double-press q replaced with a modal confirmation dialog (q → confirm y/q or cancel n/Esc)

Activity feed enrichment:

• Recent Activity events now show quest titles alongside truncated IDs for better scanability

Removed

• Alert bar: orphan/forked warning bar removed from dashboard header; diagnostic data moved to the Health section

Fixed

• Brittle test assertions: rewrote ~34 string-matching assertions across 5 test files to assert on data (IDs, counts, domain constants) instead of UI labels, section headers, and empty-state prose — eliminates false failures from vocabulary changes
• Date formatting: replaced locale-dependent toLocaleDateString() with deterministic toISOString().slice(0, 10) across all views (DashboardApp, render-status, submissions-view, lineage-view)
• Single-campaign cardinality: authorize / link now removes existing belongs-to edges before adding the new one, enforcing single-campaign assignment
• DRY violation: extracted shared statusVariant() and formatAge() into src/tui/view-helpers.ts; removed duplicate definitions from dashboard-view, submissions-view, and render-status
• Triage loop performance: hoisted WarpIntakeAdapter import and instantiation outside the per-quest loop in wizards.ts
• Quest terminology: replaced user-facing "task(s)" with "quest(s)" in backlog and deps view headers
• Array guard: guarded lines[lines.length - 1] access in landing-view against empty arrays (strict TS)
• Quit confirm overlay hidden: pressing q from landing or help screen now clears those views before showing the confirm overlay
• Non-domain terminology: renamed internal my-issues panel key to my-quests for Digital Guild consistency
• Lint violation: fixed Array<T> annotation to T[] in integration test

Added — Dashboard Enhancement Chunk 2: Components + Data

New bijou components adopted (TUI dashboard):

• separator() — section dividers replace styled-header patterns across dashboard
• badge() — status tags with visual weight in Pending Review, My Submissions, submission detail
• timeline() — Recent Activity feed with status-colored event markers
• enumeratedList() — numbered Top Blockers list
• stepper() — submission lifecycle visualization (Submitted → Reviewed → Approved → Merged/Closed)

New data surfaced in dashboard:

• Top Blockers — calls existing computeTopBlockers(), shown as numbered list (gated by deps)
• Critical Path — calls existing computeCriticalPath(), compact stat line (gated by deps)
• Blocked Quests — surfaces computeFrontier().blockedBy with waits-on detail (gated by deps)
• Submission Age — relative age in submission detail panel

Changed — BJU-002: Port render-status.ts to bijou

Replaced all legacy cli-table3 and boxen usage in the CLI status views with bijou equivalents. Zero remaining imports of either package in src/.

• 11 cli-table3 tables → bijou table()
• 6 boxen snapshot headers → bijou headerBox()
• Section headers → separator()
• Status tags → badge() with statusVariant() helper
• Manual numbered lists → enumeratedList() in deps view

Added — Second Wave: Bijou 0.10.0 Primitives

• Command palette (DashboardApp.ts): : or / opens a fuzzy-searchable commandPalette() in a modal() overlay with context-aware actions per view (claim, promote, reject, expand, approve, request-changes).
• Interactive wizards (src/cli/commands/wizards.ts): quest-wizard, review-wizard, promote-wizard, and triage CLI commands using bijou filter(), select(), input(), textarea(), and confirm() primitives.
• Integration test suite (integration.test.ts): 22 deterministic "drive" pattern tests for the full init → update → view cycle without async settling.
• Drawer detail panel (roadmap-view.ts): Quest detail now renders as a bijou drawer() + composite() overlay; DAG takes full width when no quest is selected.

Changed — NavigableTableState Migration (Phase A complete)

All three selectable views now use NavigableTableState for focus management with circular j/k navigation and built-in scroll tracking.

• Backlog view: Already migrated (first wave). Uses navigableTable() renderer.
• Submissions view: Migrated from manual selectedIndex + listScrollY to NavigableTableState. List panel now uses navigableTable() renderer.
• Roadmap view: Migrated from manual selectedIndex to NavigableTableState for focus tracking. Frontier panel rendering unchanged (custom grouped layout).
• RoadmapState.selectedIndex → RoadmapState.table: NavigableTableState
• SubmissionsState.selectedIndex / listScrollY → SubmissionsState.table
• Added rebuildRoadmapTable() and rebuildSubmissionsTable() table builders.
• selectDelta() uses navTableFocusNext/navTableFocusPrev (wrapping).
• buildPaletteItems() checks table.rows.length > 0 instead of selectedIndex >= 0.
• Lineage view retains manual selection (different interaction model with collapsible intents).

Changed — Bijou 0.10.0 Adoption

Upgraded @flyingrobots/bijou and @flyingrobots/bijou-tui from 0.6.0 to 0.10.0. Replaced hand-rolled TUI primitives with bijou builtins where strictly better.

• Overlays (overlays.ts): Rewrote confirmOverlay() and inputOverlay() to use bijou composite() + modal() — ANSI-safe cell-by-cell painting replaces manual line-splicing with box-drawing characters (~60 lines deleted).
• Status bar (DashboardApp.ts): Replaced renderStatusLine() with bijou statusBar() (~20 lines deleted).
• Toast notifications (DashboardApp.ts): Toast is now a proper composite() overlay anchored bottom-right instead of being embedded in the status bar.
• Help system (DashboardApp.ts): Deleted renderHelp() and viewHints(), replaced by auto-generated helpView() / helpShort() from keymaps — help text always stays in sync with actual bindings. All keymap builders refactored to use .group() (~50 lines deleted).
• Landing view (landing-view.ts): Manual spiral + box compositing replaced by bijou canvas() + spiralShader + composite() + modal() (~70 lines deleted).
• Spiral shader (spiral.ts): Extracted per-cell spiralShader: ShaderFn export for use with bijou canvas(), alongside the original spiralFrame().

Added — Animated Spiral Title Screen

• Spiral shader (src/tui/bijou/shaders/spiral.ts): Ported ertdfgcvb's "Spiral" ASCII shader — generates an animated full-screen character background driven by Date.now(), using the density ramp .·:;░▒▓█.
• Solid content box: Logo, copyright, and status text render inside a centered ┌─┐│ │└─┘ bordered panel with 3-char horizontal / 1-line vertical padding. The spiral fills all space outside the box.
• Full-width progress bar: Loading bar pinned to the absolute bottom row, stretching the full terminal width.
• Event loop yields (GraphContext.ts): Added setImmediate yields between the 5 heaviest fetchSnapshot() pipeline stages (syncCoverage, materialize, checkpoint, queries, neighbor resolution) so animation frames fire between CPU-heavy steps instead of freezing for the entire load.
• Backlog: task:worker-thread-loading — offload fetchSnapshot to a worker_threads Worker for true zero-hitch loading (future).

Added — Top Blockers Analysis

• computeTopBlockers() in DepAnalysis.ts: BFS-based analysis that ranks non-DONE tasks by transitive downstream impact (how many other tasks they block).
• deps CLI view: New "Top Blockers" table showing direct and transitive dependent counts per blocking task.
• Roadmap TUI view: New "Top Blockers" section in the left panel between frontier and blocked lists.

Fixed — Documentation Accuracy

• CLAUDE.md: Updated test count (60+ → 900+), added missing reopen command to reference.
• README.md: Fixed TUI view cycle names (overview/inbox → dashboard/backlog), added missing reopen command to CLI table, removed duplicate entry.

Changed — Roadmap Audit & Documentation Overhaul

• postinstall script: Added patch-package postinstall hook to package.json.
• Roadmap audit: Identified and sealed 17 quests that were DONE but still marked PLANNED/IN_PROGRESS in the graph (WVR-001–005, BJU-004–008, DSH-005, DSH-007, BX-021, SUB-TUI-001, SUB-TUI-002, SUB-REFACTOR-001, inkstatus-type-safety).
• Closed 2 stale submissions: submission:0mm4fvum740cf37 (BJU-001) and submission:0mm4nm0oa510ca6 (BJU-002) — referenced commits already on main, branches diverged past patchset snapshots.
• Bijou v0.6.0 upgrade: @flyingrobots/bijou, bijou-tui, bijou-node all upgraded from v0.5.1. New: navigableTable(), browsableList(), createInputStack(), helpView(), wizard(), filter(), dagStats(). No breaking changes.
• GRAPH_SCHEMA.md rewritten (v1.0 → v2.0.0): All 21 prefixes, 16 edge types, 10 node property contracts, edge traversal diagram, LWW rules.
• DATA_CONTRACTS.md retired: M1-era speculation superseded by GRAPH_SCHEMA.md Section 4.
• AGENT_CHARTER.md: Added DRAFT status header (unimplemented 6-agent role architecture).
• Version fixes: README (alpha.8 → alpha.11), CLAUDE.md (git-warp v12.0.0 → v12.1.0), EXECUTIVE_SUMMARY.md (339 → 500 tests), ROADMAP.md (inbox count 33 → ~100).
• README.md: Updated TUI file tree (Ink → bijou architecture).
• ROADMAP.md: Moved Weaver to DONE, added M12 Agent Protocol milestone, updated CLI Tooling scope.
• New docs: CLI-plan.md (interactive wizards, missing commands, agent protocol), docs/XYPH_Workflows.md (practical guide to all XYPH features).
• New backlog items: 20 tasks filed across CLI wizards, missing commands, agent protocol, TUI enhancements, and doc rewrites.

Fixed — Code Review

• H-1: Shared selection ordering — Extracted roadmapQuestIds, submissionIds, backlogQuestIds, lineageIntentIds into selection-order.ts. Both DashboardApp (j/k navigation) and view renderers import from the same module, preventing wrong-item-selected bugs from ordering divergence.
• M-1: dagScrollY asymmetry — Vertical DAG scroll offset now applies as an offset from auto-center, matching the existing dagScrollX behavior.
• M-2: GRAVEYARD progress exclusion — Dashboard progress bar and orphan alert exclude GRAVEYARD quests (previously inflated denominator).
• M-3: filterSnapshot submissions — filterSnapshot() now filters submissions for graveyard quests alongside scrolls and sortedTaskIds.
• M-4: "Assigned Issues" rename — Dashboard right column header changed from "My Issues" to "Assigned Issues" for clarity.
• M-7: normalizeQuestStatus hardened — All known statuses now have explicit switch cases; default branch is truly unreachable for valid input.
• M-8: Write debounce — Added writePending flag to prevent double-writes while a graph mutation is in flight. Confirm/input handlers set the flag; write-success/error clear it; write actions short-circuit when true.
• L-1/L-4: GRAVEYARD in roadmap — Added skull icon for GRAVEYARD status; excluded GRAVEYARD quests from selectable IDs in both deps and no-deps paths.
• L-3: OCP comment — Clarified that claim verification reads local state only.
• L-4: rejectQuest guard — Empty-rationale check added, matching promoteQuest.
• L-5: Clamp on reload — Per-view selectedIndex is clamped when a new snapshot arrives, preventing stale indices after quest deletion.
• L-6: Critical path phantom guard — computeCriticalPath inner loop skips dependents not in the weight map, preventing NaN propagation from phantom edges.
• L-8/N-7: Exhaustive defaults — handleViewAction and viewRenderer switch statements now have default: never guards for compile-time exhaustiveness.

Changed — Dependency Upgrade: bijou v0.5.1

• Upgraded @flyingrobots/bijou, @flyingrobots/bijou-node, and @flyingrobots/bijou-tui to v0.5.1. Resolves dual-package context issue where bijou-tui bundled a nested bijou v0.4.0 copy, causing [bijou] No default context configured crash at dashboard startup.

Changed — Performance: Query-based getQuests() and DagSource adapter

• WarpRoadmapAdapter.getQuests() now uses graph.query().match('task:*') instead of walking every node via getNodes() + per-node getNodeProps(). Reduces work from O(total_nodes) to O(task_nodes).
• Upgraded @flyingrobots/bijou to v0.5.0.
• Roadmap DAG rendering now uses a SlicedDagSource adapter backed by the existing questMap instead of building an intermediate DagNode[] array. Graph data is read lazily via the source interface — no duplication.

Changed — Topological Sort: Use git-warp Engine

• Removed manual topoSort() (Kahn's algorithm) from DepAnalysis.ts — git-warp already provides graph.traverse.topologicalSort() with the same semantics.
• GraphContext.fetchSnapshot() now computes sortedTaskIds via the engine (dir: 'in', labelFilter: 'depends-on') and includes it in GraphSnapshot.
• roadmap-view.ts reads snap.sortedTaskIds instead of calling a manual sort.
• filterSnapshot() filters sortedTaskIds when excluding GRAVEYARD quests.
• Removed 4 topoSort unit tests (coverage now lives in git-warp's own test suite).

Fixed — PR Feedback (Review Round)

• Roadmap frontier selection ordering now matches visual render order (frontier items first, then blocked, both sorted).
• Backlog write targets aligned with suggester-grouped render order, preventing wrong-target promote/reject operations.
• Toast dismissal uses tokenized expiresAt to prevent stale timers from clearing newer toasts.
• promoteQuest validates intentId is non-empty before calling IntakePort.
• Overlay dialog hint lines pad plain text before ANSI styling to fix visual width miscalculation.
• Landing progress bar width clamped to terminal width for narrow screens.
• Consolidated duplicate "### Removed" CHANGELOG sections.
• Extracted shared normalizeQuestStatus and SUBMISSION_STATUS_ORDER constants to domain entities, eliminating duplication across adapters.
• Dashboard pending-review lookup optimized from O(n) find to Map.

Changed — Phase 9: Vocabulary Rename

• INBOX → BACKLOG (suggestion pool), BACKLOG → PLANNED (vetted work).
• Read-time normalization in GraphContext (normalizeQuestStatus) ensures legacy graph data maps transparently — no graph migration needed.
• inbox-view.ts → backlog-view.ts; all references, tests, keybindings, and tab labels updated.
• IntakeService and adapters updated: sendToInbox() now writes BACKLOG; promote() transitions BACKLOG → PLANNED.
• Quest entity and status types reflect new vocabulary.
• Theme presets: INBOX tokens replaced with BACKLOG.

Added — Phase 5: Dashboard Redesign

• dashboard-view.ts: Full replacement for overview-view.ts as the default landing view.
• Project progress bar via bijou progressBar() — DONE / non-backlog ratio.
• Alert bar: warns on orphan quests (no intent) and forked patchsets.
• "In Progress" and "Pending Review" panels with top items.
• Campaign progress: per-campaign progressBar() with done/total counts.
• "Assigned Issues" panel (assigned, non-terminal quests).
• Health section: sovereignty audit ratio, orphan count, fork count.
• Graph meta section: max tick, writer count, tip SHA.
• Two-column flex layout (2:1 ratio).

Added — Phase 4: Lineage Intent Cards

• Intent descriptions surfaced from graph (description property on intent nodes, read in GraphContext).
• progressBar() completion stats per intent (DONE / total authorized quests).
• Indented quest trees under each intent with status styling.

Added — Phase 3: Interactive DAG Scrolling

• Roadmap view upgraded to bijou dagLayout() with viewport({ scrollX }).
• Auto-centers on selected node when selection changes.
• h/l keybindings for horizontal DAG scrolling.

Removed

• overview-view.ts — replaced by dashboard-view.ts.
• inbox-view.ts — replaced by backlog-view.ts.
• all-view.ts — replaced by overview-view.ts.

Added — Interactive TUI Phase 2: Review Actions + Roadmap Detail

Review actions in submissions view:

• a (approve) and x (request-changes) keybindings in submissions view.
• Input mode prompts for review comment before dispatching.
• Validates tip patchset exists — shows error toast if missing.
• reviewSubmission() write command factory with auto-generated review IDs.
• SubmissionPort wired into WriteDeps and DashboardDeps.

Roadmap detail panel:

• Third flex column (28 cols) appears when a quest is selected.
• Shows: quest ID, title, status, hours, owner, campaign title, intent, dependency count with status icons, submission status, and scroll info.
• Hidden when no quest is selected — DAG takes full width.

DAG selected node highlighting:

• Selected quest node in DAG uses primary theme token, overriding the critical-path warning color for visual distinction.

Tests:

• 8 new tests: review keybindings (a/x enter input mode), error toast on missing tip patchset, review input flow, detail panel content/deps/ submission status, hidden detail panel.

Added — Interactive TUI Phase 1: New Views, Selection, and Write Operations

New views:

• overview-view.ts: Summary dashboard with quest status counts, submission status counts, health metrics (sovereignty audit, orphan quests, forked patchsets), campaign list, and graph meta — replaces the raw-table all-view.
• submissions-view.ts: Master-detail layout (35/65 flex split). Left panel shows submissions sorted by status priority (OPEN first); right panel shows expanded detail with patchset chain, reviews, and decisions.

Selection and navigation:

• Per-view state: roadmap.selectedIndex, submissions.selectedIndex, inbox.selectedIndex with j/k (or arrow keys) for navigation.
• Roadmap frontier panel highlights selected quest with ▶ indicator.
• Inbox table highlights selected item with ▶ indicator.
• Submissions list highlights selected entry; Enter toggles detail expansion.
• View-specific hint bar shows available keybindings per view.

Write operations:

• write-cmds.ts: Cmd factories for claimQuest (OCP via direct graph patch), promoteQuest (IntakePort), rejectQuest (IntakePort).
• Confirm mode: c on roadmap → confirm dialog (y/n) → claim quest → toast.
• Input mode: p on inbox → text input for intent ID → promote quest → toast. d on inbox → text input for rationale → reject quest → toast.
• overlays.ts: Centered confirm dialog and text input overlays rendered over view content.
• Toast notifications in status line (success=green, error=red, auto-dismiss 3s).

Architecture:

• DashboardModel expanded: per-view state objects, interaction modes (normal/confirm/input), PendingWrite action type, toast state.
• DashboardMsg expanded: write-success, write-error, dismiss-toast.
• DashboardDeps expanded: graphPort: GraphPort for direct graph patches.
• Tab order: roadmap → submissions → lineage → overview → inbox (5 views).
• View-specific keymaps via separate KeyMap instances per view.

Tests:

• 57 new tests across DashboardApp.test.ts and views.test.ts: 5-view cycling, per-view state init, selection (j/k), confirm mode, input mode, toast lifecycle, submission expand/collapse, overview metrics, submissions detail rendering.

Changed — BJU-002: Port TUI Views to Bijou Components

• Ported 4 dashboard views from stub placeholders to full rendering logic using bijou headerBox(), table(), and tree() components.
• roadmap-view.ts: quests grouped by campaign with table().
• inbox-view.ts: INBOX quests grouped by suggestedBy with per-suggester table().
• all-view.ts: conditional sections (campaigns, intents, quests, scrolls, approvals) each with table().
• lineage-view.ts: intent→quest→scroll hierarchy using tree() with scroll marks (sealed/unsealed) and orphan quest detection.

Added

• 20 new view tests in views.test.ts covering null/empty/populated snapshots, orphan quests, scroll marks, truncated rejection rationale.
• CI strict-policy job: grep-based gate rejecting eslint-disable, @ts-ignore, @ts-expect-error, and @ts-nocheck in all source files.
• ESLint noInlineConfig: true — inline lint bypass comments are now a hard error.

Fixed

• lineage-view.ts: orphan quests now shown even when no intents exist (previously hidden by early return).
• all-view.ts: node total now includes submissions, reviews, and decisions.

1.0.0-alpha.10 - 2026-02-26

Changed — BJU-001: Bijou TUI Migration (Theme Bridge + TEA App Shell)

Framework: Ink/React → bijou-tui TEA architecture

• Replaced the Ink/React Dashboard.tsx component tree with a pure-function TEA (The Elm Architecture) app powered by @flyingrobots/bijou-tui's run().
• New DashboardApp.ts: init() / update() / view() with immutable DashboardModel, typed DashboardMsg union, and Cmd<DashboardMsg> effects.
• Keybindings via bijou-tui's createKeyMap() — q, Tab, Shift+Tab, r, ?.
• Layout via flex() (column direction) with tabs() tab bar from @flyingrobots/bijou.

Theme: local token system → bijou theme bridge

• Replaced tokens.ts, presets.ts, resolve.ts, chalk-adapter.ts, gradient.ts, and ink-adapter.tsx with bijou's theme system.
• New xyph-presets.ts: extends bijou's CYAN_MAGENTA and TEAL_ORANGE_PINK with 15 XYPH status keys + intentHeader UI key via extendTheme().
• New bridge.ts: lazy-init singleton that configures bijou's global context with XYPH presets, env-var theme selection (XYPH_THEME), and NO_COLOR support.
• theme/index.ts barrel: init-guarded styled() / styledStatus() wrappers that call ensureXyphContext() before delegating to bijou.

Views: stub views for BJU-002 migration

• Five new bijou view functions (roadmap-view.ts, lineage-view.ts, all-view.ts, inbox-view.ts, landing-view.ts) with summary stats and placeholder text for full rendering in BJU-002.

Added

• @flyingrobots/bijou ^0.2.0, @flyingrobots/bijou-node ^0.1.0, @flyingrobots/bijou-tui ^0.1.0 as dependencies.
• git-warp API reference section in CLAUDE.md.

Removed

• 17 old Ink/React files: Dashboard.tsx, GraphProvider.tsx, HelpModal.tsx, QuestDetailPanel.tsx, Scrollbar.tsx, StatusLine.tsx, and all 5 view components (AllNodesView.tsx, InboxView.tsx, LandingView.tsx, LineageView.tsx, RoadmapView.tsx), plus 6 local theme modules.
• chalkFromToken() — replaced by bijou's styled().

Fixed — Code Review (9 issues resolved: 3 high, 4 medium, 2 low)

• High: Ctrl+C was swallowed in landing (while loading) and help modes — hoisted Ctrl+C handler above mode-specific branches so it always quits.
• High: q did nothing in help mode despite help text listing it as quit — added q → quit handling in help mode to match displayed shortcuts.
• High: Lint violations (explicit-function-return-type on makeApp(), no-non-null-assertion on resolver!) — added return type annotation and replaced ! with explicit guard + throw.
• Medium: initialized flag set before createBijou()/createThemeResolver() completed — moved flag after success to prevent permanently broken state on init failure.
• Medium: Refresh race condition — added monotonic requestId to model and snapshot messages; stale responses from superseded fetches are now discarded.
• Medium: styled()/styledStatus() re-exported without init safety — replaced raw re-exports with wrappers that call ensureXyphContext() before delegating.
• Medium: _resetBridgeForTesting() did not reset bijou's global default context — now calls _resetDefaultContextForTesting() for full test isolation.
• Low: Dead code removed — dismiss-landing message type/handler, unreachable dismiss action, unused wordmarkText field threaded through model and deps.
• Low: Landing "Press any key" prompt shown during loading (when keys are ignored) — prompt now only appears after data loads or on error.

1.0.0-alpha.9 - 2026-02-25

Changed — Shared Graph Architecture & GraphContext Refactor

Architecture: one WarpGraph instance per process

• Introduced GraphPort (port) and WarpGraphAdapter (infrastructure) — a process-wide singleton for the shared WarpGraph instance. All adapters receive it via dependency injection instead of creating their own WarpGraphHolder.
• Rewired WarpSubmissionAdapter, WarpIntakeAdapter, WarpRoadmapAdapter, coordinator-daemon, xyph-actuator, and xyph-dashboard to use GraphPort DI.
• Eliminated WRITER_CAS_CONFLICT errors caused by multiple WarpGraphHolder instances sharing the same writerId.

Architecture: kill the adapter-walks-every-node anti-pattern

• Replaced monolithic WarpDashboardAdapter (542 LoC) + DashboardService (113 LoC) with GraphContext — a single shared gateway using graph.query() for typed node fetching and graph.traverse for graph algorithms.
• New GraphProvider.tsx (React context) delivers the GraphContext to TUI components.
• Extracted DepAnalysis.ts — pure domain functions for frontier detection and critical-path DP, replacing the algorithmic parts of WeaverService.

Performance: atomic graph.patch() for all writes

• Converted all adapters and xyph-actuator.ts from manual syncCoverage() + materialize() + createPatchSession() + commit() to graph.patch(p => { ... }).
• Eliminated redundant materialize() and syncCoverage() calls — autoMaterialize: true makes writes immediately visible to reads on the same instance.
• Submission lifecycle integration test: 15s timeout → default 5s, actual runtime ~1.2s.

Fixed

• GraphContext cache invalidation: replaced hasFrontierChanged() (only detects external patches) with frontier key comparison that catches both in-process graph.patch() writes and external mutations from syncCoverage().

Removed

• WarpGraphHolder — replaced by WarpGraphAdapter.
• WarpDashboardAdapter, DashboardService, DashboardPort — replaced by GraphContext.
• WeaverService, WeaverPort, WarpWeaverAdapter — replaced by DepAnalysis + direct graph.traverse calls.
• WarpDashboardAdapter.test.ts, DashboardService.test.ts, WeaverService.test.ts, WarpWeaverAdapter.test.ts — tests migrated to DepAnalysis.test.ts and existing integration suites.

Code review — 7 issues resolved (1 critical, 2 high, 2 medium, 2 low)

• Critical: invalidateCache() no longer calls graphPort.reset() — previous behavior orphaned the shared graph and violated the singleton invariant. Now only clears GraphContext's own cached snapshot and frontier key.
• High: Removed inconsistent syncCoverage() call from the depend command — no other write command calls it post-refactor.
• High: Inlined dead getGraph() wrapper in xyph-actuator.ts — all ~7 call sites now use graphPort.getGraph() directly; removed unused WarpGraph type import.
• Medium: Reduced getStateSnapshot() calls in fetchSnapshot() from 3 to 2 — early call for cache check, post-materialize call for graphMeta and cached frontier key.
• Medium: batchNeighbors() now uses Promise.all instead of Promise.allSettled — neighbor resolution errors surface immediately instead of being silently swallowed.
• Low: Batched separate graph.patch() seed calls into single patches in WarpIntakeAdapter.test.ts and WarpSubmissionAdapter.test.ts.
• Low: Fixed JSDoc in GraphPort.ts — replaced reference to private _onPatchCommitted with public description.

1.0.0-alpha.8 - 2026-02-25

Milestone 7: Weaver — Task Dependency Graph

Added

• depend <from> <to> CLI command: declares depends-on edges between task: nodes with cycle detection, self-dependency prevention, and existence validation.
• status --view deps: new dashboard view showing frontier (ready tasks), blocked tasks, topological execution order, and critical path with total hours.
• WeaverService (domain): pure validation and computation — frontier detection, topological ordering (via git-warp v12 LogicalTraversal), critical path DP over dependency DAG.
• WeaverPort (port): write interface for dependency edge persistence.
• WarpWeaverAdapter (infrastructure): implements both WeaverPort and WeaverReadModel using git-warp v12's graph.traverse.isReachable(), graph.traverse.topologicalSort(), and graph.getEdges().
• QuestNode.dependsOn field in dashboard model; WarpDashboardAdapter reads depends-on edges during snapshot construction.
• renderDeps() renderer: frontier table, blocked-tasks table, numbered execution order, and critical path chain display.
• CI workflow (.github/workflows/ci.yml): enforces build, lint, and test gates on every PR and push to main as three parallel jobs.
• CI audit job: runs npm audit --omit=dev to catch known vulnerabilities in production dependencies.
• 31 new tests (23 unit + 8 integration), 369 total.

Fixed

• inspect-graph.ts: replaced unsafe inline type cast with toNeighborEntries() runtime guard.

1.0.0-alpha.7 - 2026-02-24

TUI Overhaul: Fullscreen, Flicker-Free Rendering & Responsive Layout

Added

Alternate screen + flicker-free rendering shim

• xyph-dashboard.tsx: enters alternate screen buffer (\x1b[?1049h) on launch, restores on exit.
• Patches stdout.write to replace Ink's clearTerminal (erase + rewrite) with cursor-home + erase-to-EOL, eliminating full-screen flash on every render.
• Proper cleanup on SIGINT/SIGTERM — always restores the original terminal.

git-warp verbose logging in the gutter

• TuiLogger (src/tui/TuiLogger.ts): new LoggerPort implementation that captures git-warp internal logs via a callback for display in the TUI.
• WarpGraphHolder and WarpDashboardAdapter accept optional LoggerPort, passed to WarpGraph.open().
• Dashboard subscribes to logger with 150ms throttle to prevent rapid re-renders.

Persistent gutter on all screens

• StatusLine now renders on every screen (landing, loading, error, main views) — no more early returns that skip it.
• Dashboard restructured: root <Box height={rows}> wrapper with flexGrow={1} content area and pinned gutter.
• LandingView uses flexGrow={1} + justifyContent="center" for vertical centering within the flex container.

Changed

Status line simplified

• Replaced tick: 142 (8a062e1) | me: 122 | writers: 5 with t=142 — one number, the global frontier tick.
• Dropped per-writer tick and writer count (implementation details, not user-facing).
• Gutter log line prefixed with [warp(t=N)] showing the current graph tick when available.

Cold start performance

• Removed syncCoverage() from first-load path — WarpGraph.open() discovers refs automatically.
• Restored syncCoverage() on refresh path — discovers external mutations committed since last materialize.
• Added createCheckpoint() after materialize() — persists materialized state so subsequent launches load from checkpoint instead of replaying all patches.
• tipSha now derived from createCheckpoint() return value (content hash of materialized state) instead of writer tip commit.

Responsive full-width table layout

• All four views (RoadmapView, LineageView, AllNodesView, InboxView) now calculate column widths dynamically from terminal width.
• Title columns absorb remaining width instead of using hardcoded .padEnd() values.
• Tables fill edge-to-edge on any terminal size (min 12 chars for title).

Fixed

• Ghost content on title screen: erase-to-end-of-line (\x1b[K) injected after each line during cursor-home redraws, preventing remnants from longer previous renders.
• Full terminal height: root Box in Dashboard and LandingView now properly fills terminal via height={rows} / flexGrow={1}.

Code review — 15 issues resolved (4 HIGH, 6 MEDIUM, 5 LOW)

• High: TuiLogger.child() now delegates onEntry through parent chain — children created before onEntry is set no longer get a permanent null callback.
• High: HelpModal renders as the content area (replacing views) instead of appended below StatusLine — no longer overflows height={rows}.
• High: WarpDashboardAdapter calls syncCoverage() on refresh path (when cachedSnapshot !== null) to discover external mutations; skipped on first load where WarpGraph.open() handles discovery.
• High: xyph-dashboard.tsx cleanup guard prevents double terminal-restore on SIGINT → exit handler chain.
• Medium: StatusLine always renders 2 lines (empty second line when no logLine) — stabilizes gutterLines constant.
• Medium: Log prefix changed from unstable [warp(SHA)] to [warp(t=N)] using graph tick.
• Medium: LineageView trailing spaces (+4) added to lineageFixedW; status text padded to fixed 14 chars.
• Medium: AllNodesView magic 10 extracted to scrollSealW constant with comment.
• Medium: package.json version bumped to 1.0.0-alpha.7.
• Low: TuiLogger.emit() nested ternary refactored to explicit if/else.
• Low: InboxView suggestedBy uses suggestedByW - 2 instead of hardcoded 14.
• Low: AllNodesView questSuffixW components documented with inline comments.
• Low: LineageView intentIdW reverted from 32 to 30 (matches original behavior).
• Low: Dashboard.tsx cols moved inside else branch where it's actually used.

Codex review — 2 issues resolved (1 P1, 1 P2)

• P1: WarpDashboardAdapter syncCoverage() on refresh path — already addressed in H-3 above.
• P2: xyph-dashboard.tsx SIGINT/SIGTERM handlers now use conventional signal exit codes (130/143) instead of exit(0), so shell wrappers and supervisors can distinguish cancel/kill from success.

CodeRabbit review — 6 issues resolved (1 major, 3 minor, 2 nit)

• Major: WarpDashboardAdapter.createCheckpoint() wrapped in try/catch — checkpoint failures no longer block snapshot rendering; falls back to tipSha='unknown'.
• Minor: CHANGELOG.md unused [Unreleased] link reference resolved by adding ## [Unreleased] section header (MD053).
• Minor: AllNodesView sealedBy column now truncated before padding to prevent line wraps on long values.
• Minor: LandingView layout — full test suite (npm run build + npm run test:local) verified: 338/338 pass, lint clean.
• Nit: StatusLine exports STATUS_LINE_HEIGHT constant; Dashboard.tsx imports it instead of hardcoding gutterLines = 2.
• Nit: TuiLogger.onEntry setter throws on child loggers — prevents silent misconfiguration since resolveOnEntry() always defers to root.

---

1.0.0-alpha.6 - 2026-02-22

Dashboard Performance & UX Improvements

Changed

Performance — parallel graph snapshot loading

• WarpDashboardAdapter.fetchSnapshot(): batch-fetch all node props via Promise.all instead of sequential awaits (142 nodes).
• Pre-fetch all outgoing neighbors() calls in a single parallel batch, eliminating sequential I/O across 4 passes.
• buildSubmissionData() converted from async to synchronous — uses pre-fetched neighbors cache instead of live graph queries.

Loading UI — activity log on landing screen

• DashboardPort.fetchSnapshot() now accepts optional onProgress callback for phase-level progress reporting.
• LandingView displays a live activity log (last 10 lines) during graph loading, replacing the static "Loading…" message.

Copywriting — brand consistency

• "WARP GRAPH STATUS" → "XYPH GRAPH STATUS" on the landing screen.
• "Loading WARP graph snapshot…" → "Loading project graph snapshot…".

Fixed

Tab cycling — defensive key guards

• Added explicit return after key.tab handler in Dashboard.tsx to prevent fall-through.
• All 4 view components (RoadmapView, LineageView, AllNodesView, InboxView) now explicitly ignore key.tab at the top of their useInput handlers, ensuring Tab keypresses are cleanly handled only by Dashboard's view-switching logic.

Code review — 31 issues resolved across 3 rounds

• High: Removed getTheme/styled TUI imports from domain-layer TriageService and SovereigntyService — domain services now use plain console.log/console.warn (hexagonal architecture fix).
• High: Changed 'Opening WARP graph…' → 'Opening project graph…' in WarpDashboardAdapter progress log (brand consistency).
• High: WarpDashboardAdapter batch prop fetch changed from Promise.all → Promise.allSettled for partial-failure resilience.
• High: WarpDashboardAdapter batch neighbor fetch changed from Promise.all → Promise.allSettled (mirrors props pattern).
• Medium: Added missing space after comma in 91 styled() calls in xyph-actuator.ts.
• Medium: Removed redundant syncCoverage() + materialize() from WarpGraphHolder.initGraph().
• Medium: TriageService.linkIntent() log moved after null guard to avoid misleading entries.
• Medium: StatusLine prefix /// WARP → /// XYPH; remaining WARP branding in STYLE_GUIDE updated.
• Low: ThemeProvider now uses useMemo for stable context value (prevents unnecessary re-renders).
• Low: Exhaustive never check added to chalkFromToken TextModifier switch.
• Low: NaN guard added to gradient.ts for duplicate stop positions.
• Low: WarpDashboardAdapter uses NeighborEntry type instead of inline shape.
• Low: result.reason safely stringified in batch warning logs.
• Low: Test hygiene — warnSpy wrapped in try/finally, hardcoded hex replaced with token lookup, vi.unstubAllEnvs() for env restoration, satisfies assertion on status keys, status keys derived from theme.
• Low: Dashboard.tsx now clears loadLog state when snapshot loads.
• Low: Replaced inline token construction in coordinator-daemon.ts.
• Low: Re-indented scripts/bar-demo.ts (project convention).
• Low: CHANGELOG structure — merged [Unreleased] into [1.0.0-alpha.6], added comparison link.

---

Theme Token System — Full Visual Layer Migration

Added

Theme module (src/tui/theme/)

• New tokens.ts: RGB, GradientStop, TextModifier, TokenValue, InkColor, StatusKey, Theme type definitions. All colors stored as #RRGGBB hex strings for deterministic cross-terminal rendering.
• New presets.ts: CYAN_MAGENTA theme (matches all prior hardcoded values exactly) and TEAL_ORANGE_PINK theme (new candidate palette from gradient experiment). Helper tv() for concise token definition.
• New gradient.ts: lerp3() N-stop linear interpolation extracted from scripts/bar-demo.ts.
• New resolve.ts: isNoColor() (per no-color.org spec), getTheme() singleton with XYPH_THEME env var selection, resolveTheme() for React context, _resetThemeForTesting().
• New chalk-adapter.ts: chalkFromToken(), styled(), styledStatus() — chalk from theme tokens with NO_COLOR support (hex skipped, modifiers preserved).
• New ink-adapter.tsx: ThemeProvider React context component, useTheme() hook with singleton fallback for incremental migration.
• New index.ts: barrel re-exports.
• Theme selection via XYPH_THEME env var (e.g., XYPH_THEME=teal-orange-pink).
• NO_COLOR respected: ink() returns undefined → Ink renders default terminal color; chalkFromToken() skips .hex() → only modifiers apply.

Tests — 44 new tests (338 total, up from 249)

• gradient.test.ts — 9 tests: boundary values, mid-stop interpolation, single-stop and empty-stop edge cases.
• presets.test.ts — 13 tests: all status keys defined in both presets, hex format validation, gradient stop ordering.
• resolve.test.ts — 11 tests: theme selection, NO_COLOR detection, singleton caching, unknown theme fallback with warning.
• chalk-adapter.test.ts — 11 tests: styled output, status rendering, modifier application, NO_COLOR mode.

Changed

TUI components — color literals → theme tokens (10 files)

• Scrollbar.tsx: cyan/gray → ui.scrollThumb/scrollTrack.
• HelpModal.tsx: cyan border → border.primary, yellow headings → semantic.warning.
• QuestDetailPanel.tsx: dropped STATUS_COLOR import, uses inkStatus() and semantic tokens.
• Dashboard.tsx: cyan/gray tabs → ui.cursor/semantic.muted, yellow/red states → semantic tokens.
• LandingView.tsx: green progress → semantic.success, cyan logo → ui.logo, yellow milestone → semantic.warning.
• RoadmapView.tsx: cyan cursor → ui.cursor, blue campaign headers → ui.sectionHeader, status lookups via inkStatus().
• LineageView.tsx: magenta intent headers → ui.intentHeader, cyan cursor → ui.cursor.
• AllNodesView.tsx: green section headers → semantic.success, cyan cursor → ui.cursor.
• InboxView.tsx: magenta headers → ui.intentHeader, cyan/yellow/red modal borders → border.*.
• xyph-dashboard.tsx: wraps <Dashboard> in <ThemeProvider>.

CLI consumers — chalk → theme tokens (6 files)

• xyph-actuator.ts: all 91 chalk.* calls replaced with styled(getTheme().theme.semantic.*, ...).
• render-status.ts: eliminated both local STATUS_COLOR maps; all chalk.* calls → styled()/styledStatus().
• coordinator-daemon.ts: 10 chalk.* calls → styled().
• inspect-graph.ts: 7 chalk.* calls → styled().
• TriageService.ts: chalk.cyan → styled(semantic.info, ...) — also fixes hexagonal architecture violation (domain layer no longer imports chalk directly).
• SovereigntyService.ts: chalk.yellow → styled(semantic.warning, ...) — same hexagonal fix.

Gradient integration

• scripts/bar-demo.ts: imports lerp3 from src/tui/theme/gradient.js and gradient presets from src/tui/theme/presets.js; removed local duplicate definitions.

Removed

• src/tui/status-colors.ts — replaced by theme.status.* tokens. Both the TUI StatusColor map and the CLI STATUS_COLOR function map are now unified in the theme presets.

---

Backlog Reconciliation & Roadmap Triage

Added

• New intent:CLI-FOUNDATION — sovereign intent for CLI tooling (identity, packaging, time-travel, ergonomics).
• New campaign:CLITOOL (Milestone 10: CLI Tooling) — 20 quests promoted from inbox.

Changed

• VISION_NORTH_STAR.md upgraded from v1.1.0 to v1.2.0: added history-first computing, stigmergic workflows, Git-as-settlement layering, expanded planning compiler and end state sections.
• Deleted VISION_NORTH_STAR_v1.2_draft.md (promoted to authoritative).
• README milestones table: Milestone 6 (SUBMISSION) marked DONE, Milestone 10 (CLI TOOLING) added.
• Full inbox triage: 31 items processed — 1 rejected (SUB-ENTITY-001, fixed in PR #10), 20 promoted to campaign:CLITOOL, 10 promoted to campaign:DASHBOARD.

---

Milestone 6: Native WARP Graph Submission & Review Workflow

Added

Submission lifecycle (SUB-001)

• New graph node types: submission:, patchset:, review: prefixes added to schema.
• New edge types: submits, has-patchset, supersedes, reviews, decides.
• Submission and Patchset domain entities with constructor validation (src/domain/entities/Submission.ts).
• Three pure computed functions: computeTipPatchset() (head selection from supersedes chain, fork detection), computeEffectiveVerdicts() (latest-per-reviewer, comment exclusion), computeStatus() (5-state derivation: OPEN → CHANGES_REQUESTED → APPROVED → MERGED | CLOSED).
• SubmissionService domain validation service with validateSubmit, validateRevise, validateReview, validateMerge, validateClose. Error codes: [FORBIDDEN], [INVALID_FROM], [NOT_FOUND], [MISSING_ARG], [CONFLICT], [AMBIGUOUS_TIP].

Ports & adapters (SUB-002)

• SubmissionPort interface: submit(), revise(), review(), decide() — graph-only persistence, caller-generated IDs.
• WorkspacePort interface: getWorkspaceRef(), getCommitsSince(), getHeadCommit(), isMerged(), merge() — VCS abstraction (Git today, JIT tomorrow).
• WarpSubmissionAdapter: implements both SubmissionPort (write) and SubmissionReadModel (read) against the WARP graph.
• GitWorkspaceAdapter: implements WorkspacePort using local git plumbing.

CLI commands (SUB-003)

• xyph submit <quest-id> --description "..." — creates submission + first patchset, auto-fills workspace/head/commits from git.
• xyph revise <submission-id> --description "..." — adds new patchset superseding current tip.
• xyph review <patchset-id> --verdict approve|request-changes|comment --comment "..." — posts a review.
• xyph merge <submission-id> --rationale "..." — validates APPROVED status, performs git settlement, creates merge decision, auto-seals quest (scroll + GuildSeal + DONE).
• xyph close <submission-id> --rationale "..." — creates close decision without merging.

Dashboard (SUB-004)

• SubmissionNode, ReviewNode, DecisionNode view models added to dashboard.ts.
• QuestNode gains optional submissionId field.
• WarpDashboardAdapter classifies submission/patchset/review/decision nodes and computes status using domain functions.
• New renderSubmissions() renderer; wired as --view submissions.
• Submission status colors: OPEN (cyan), CHANGES_REQUESTED (yellow), APPROVED (green), MERGED (green), CLOSED (dim).

Integration safeguards (SUB-005)

• seal command now warns if a non-terminal submission exists for the quest, suggesting merge instead.
• seal and merge are independent paths to quest DONE — both remain valid.

Changed

Submission workflow jank fixes (6 items)

• WarpSubmissionAdapter: replaced O(n) full-graph scans with graph.neighbors(nodeId, 'incoming', edgeLabel) edge traversal in getOpenSubmissionsForQuest, getPatchsetRefs, getReviewsForPatchset, getDecisionsForSubmission.
• WarpDashboardAdapter: extracted 170-line fourth pass (submission/review/decision assembly) into private buildSubmissionData() helper, reducing fetchSnapshot() from ~390 to ~240 lines.
• GitWorkspaceAdapter.merge(): saves current branch before checkout, restores in finally block to avoid silently switching the user's working branch.
• generateId(): zero-padded base36 timestamp to 9 chars (17-char fixed-length IDs), now lexicographically sortable by creation time.
• PatchsetRef.supersedes renamed to supersedesId across all files for clarity.
• Added comments documenting the decision: prefix collision guard (shared between old concept/decision nodes and new submission decisions; type === 'decision' discriminator already handles it).

Tests — 63 new tests (249 total)

• test/unit/Submission.test.ts — 29 tests: entity constructors, computeTipPatchset (linear chain, forked heads, tie-breaking), computeEffectiveVerdicts (latest-per-reviewer, comment exclusion), computeStatus (all 5 rules, custom thresholds).
• test/unit/SubmissionService.test.ts — 24 tests: all validate* methods with error code coverage.
• test/integration/WarpSubmissionAdapter.test.ts — 6 tests: submit, revise, review, decide(close), full lifecycle (submit → request-changes → revise → approve → merge), getOpenSubmissionsForQuest terminal exclusion.

1.0.0-alpha.5 - 2026-02-20

Milestone 5: WARP Dashboard TUI Overhaul

Added

Landing Screen (TUI-001)

• New LandingView component displayed on startup before the main dashboard.
• Shows a randomly selected compact ASCII logo (logos 2, 3, 7, 8, 9, 10 — ≤15 lines each).
• Stats panel: ASCII progress bar (█░ per-quest completion), current active milestone, next 3 BACKLOG/PLANNED quests.
• Footer: copyright + GitHub link. Hint: any key to continue · q to quit.
• xyph-dashboard.tsx reads logo file at launch via readFileSync; falls back to 'XYPH' on error.

Help Modal (TUI-002)

• New HelpModal component: full-screen cyan border overlay accessible from any view via ?.
• Contains XYPH glossary (Quest, Intent, Campaign, Scroll, Seal, Guild, WARP) and complete key bindings table.
• Closes on Esc or ?.

Quest Detail Panel (TUI-003)

• New reusable QuestDetailPanel component showing full quest metadata: ID, title, status (colored), hours, agent, campaign (with title lookup), intent (with title lookup), scroll ✓, completedAt, suggestedBy/At, rejection history, reopen history.
• Used by RoadmapView and AllNodesView detail modals.

RoadmapView: Fold/Unfold + Detail Modal (TUI-004)

• Unified selectedVIdx navigates both campaign headers and quest rows with ↑↓.
• Space on a campaign header folds/unfolds that milestone; header shows ▶ (folded) or ▼ (expanded).
• Space on a quest opens a full-screen QuestDetailPanel modal; Esc closes it.

AllNodesView: Quest Selection + Detail Modal (TUI-006)

• ↑↓ now navigates through quest rows only (other node types are display-only).
• Space on a selected quest opens a full-screen QuestDetailPanel modal; Esc closes it.
• Status indicator text updated to include quest position and Space: quest detail hint.

InboxView: Gmail-Style Rework (TUI-007)

• Replaced fixed DETAIL_LINES = 10 constant with proportional 40%/60% split: 40% list, 60% detail.
• List rows enriched: ▶ indicator, ID, title, suggestedBy, date, ↩ reopened marker.
• Detail pane always visible (no toggle needed); shows full inbox lifecycle fields.

Logo Loader + XYPH Wordmark (TUI-008)

• New src/tui/logo-loader.ts utility module: selectLogoSize() picks small/medium/large based on terminal dimensions; loadRandomLogo() reads .txt files from the reorganized logos/{family}/{size}/ directory structure, trims leading and trailing blank lines, and falls back to plain 'XYPH' on error.
• Dimension-aware logo selection: logos are filtered by actual width/height against terminal constraints before random pick. If nothing in the preferred size fits, cascades down (large → medium → small) automatically.
• XYPH wordmark rendered in the upper-right corner of the dashboard header (dimmed). Hidden on narrow terminals (< 50 cols).
• All four views (RoadmapView, LineageView, AllNodesView, InboxView) accept an optional chromeLines prop computed from actual header height, replacing the hardcoded CHROME_LINES = 4 constant.

Fixed

LineageView: INBOX Bug Fix + Selection (TUI-005)

• INBOX quests no longer appear in the orphan ("sovereignty violation") list — they haven't been promoted yet and genuinely lack an intent.
• Added selectedVIdx state with ▶ indicator highlighting the selected quest row.

TUI layout stability (TUI-009)

• Landing page logo centered as a single block instead of per-line centering — multi-width ASCII art lines no longer scatter horizontally.
• Dashboard header uses alignItems="flex-start" and splits tab labels / hint text onto separate rows — wordmark position is now stable across all views.
• LineageView intent-header, scroll-sub, and orphan rows now truncate long text to prevent terminal line wrapping that pushed the header off-screen.
• Quest detail modal (RoadmapView, AllNodesView) renders inside a fixed-height wrapper matching the normal list height — opening/closing the modal no longer causes layout shifts.

Lint compliance — 28 errors resolved

• Replaced Array<T> with T[] syntax across 5 files (dashboard.ts, CoordinatorService.ts, GuildSealService.ts, WarpRoadmapAdapter.ts, RoadmapPort.ts).
• Converted OrchestrationFSM from static-only class to exported const object with standalone functions (no-extraneous-class).
• Removed as any casts: ed25519 hashes.sha512 polyfill now uses typed API directly; loadKeyring uses proper Record<string, unknown> narrowing instead of as any; ajv-formats/ajv-errors use typed CJS interop pattern.
• Replaced non-null assertions (!) with proper guards in validatePatchOps.ts array access and crypto.ts canonicalize.
• Added missing return types on 4 functions (crypto.ts sha512, signPatchFixture.ts sha512/generateTestKeypair, validatePatchOps.ts CLI IIFE).
• Removed redundant type annotation (no-inferrable-types) in RebalanceService constructor.
• Removed stale eslint-disable no-console directive from validatePatchOps.ts.
• Added "own every failure" policy to CLAUDE.md — agents must fix broken things they encounter, never dismiss errors as pre-existing.

Code review — 5 CodeRabbit findings resolved

• Bug: OrchestrationFSM.transitionToNormalize eventId now uses injected context.clock (now) instead of new Date(), fixing non-deterministic date fragments under frozen-time test scenarios.
• Nit: minimatch override tightened from >=10.2.1 to ^10.2.1 — prevents accidental major version jumps.
• Nit: AjvPlugin type widened to accept optional options arg and return Ajv instance (matches actual ajv-formats/ajv-errors Plugin signature).
• Nit: sha512 polyfill exported from crypto.ts and imported in signPatchFixture.ts — eliminates copy-paste duplication and redundant createHash import.
• Nit: computeDigest parameter narrowed from Record<string, unknown> to Record<string, Json> — removes unsafe as Json cast, surfacing type constraint at call site.

Changed

README aligned with canonical documentation

• Replaced "Causal Operating System for Agentic Orchestration" tagline with "The Planning Compiler for Agentic Coordination" per VISION_NORTH_STAR.md.
• Rewrote "How XYPH Works (Part I)" — removed informal GitHub comparison, added the Planning Compiler paradigm (Source → IR → Target) and Agentic Coordination Problem framing. Added LWW conflict resolution mention.
• Added "How XYPH Works (Part II)" sections: Digital Guild Model (Quests, Campaigns, Intents, Scrolls, Guild Seals, Genealogy of Intent), Planning Pipeline (Mermaid state diagram with fail-closed/ROLLED_BACK paths), and Policy Engine (MUST/SHOULD/COULD three-tier table).
• Constitution section expanded from 2 articles to all 4: Law of Determinism (Art. I), Law of DAG Integrity (Art. II), Law of Provenance (Art. III), Law of Human Sovereignty (Art. IV).
• Canonical Docs listing expanded from 5 to all 21 documents, organized into 6 categories (Vision & Governance, Architecture & Pipeline, Data & Schema, Security & Audit, Quality & Policy, RFCs).

README rewritten with progressive-disclosure walkthrough

• Replaced flat "Core Concepts / For Humans / For Agents" structure with a narrative walkthrough: Ada (human) and Hal (agent) build a feature together, introducing domain vocabulary (Intent, Quest, Campaign, Scroll, Guild Seal, OCP, Genealogy of Intent) inline on first use.
• Dashboard keybindings and CLI commands moved to compact reference tables after the walkthrough.
• Expanded Constitution section with links to all canonical specifications.
• CLAUDE.md updated: replaced stale "Current Status" with actuator-based planning workflow and full command reference.

Added CONTRIBUTING.md

• Development workflow, quality gates, Constitution summary, and full command reference table.

Security

minimatch ReDoS vulnerability (CVE: GHSA-3ppc-4f35-3m26)

• Added npm overrides to force all transitive minimatch instances to ^10.2.1, patching a Regular Expression Denial of Service (ReDoS) vulnerability where patterns with many consecutive * wildcards cause exponential backtracking (O(4^N)).
• All six affected instances (via eslint, eslint-plugin-import, @eslint/config-array, @eslint/eslintrc, @typescript-eslint/typescript-estree, glob) now resolve to minimatch@10.2.2.

ajv ReDoS vulnerability (GHSA-2g4f-4pwh-qvx6)

• Bumped transitive ajv (via @eslint/eslintrc, eslint) to patched versions via npm audit fix.

Dependency upgrade

• Upgraded @git-stunts/git-warp from 11.3.3 to 11.5.0.

Code Review — 68 issues resolved (CR-001)

• Critical: IngestService rewritten — task: prefix guard, new Quest() in try/catch (skips invalid lines), clean formatting.
• Critical: package-lock.json synced to alpha.4.
• High: Eliminated actuator TOCTOU — promote/reject/reopen now call WarpIntakeAdapter directly instead of dual-graph validate-then-write.
• High: Added reopen() to IntakePort + WarpIntakeAdapter; all intake methods now return commit SHA.
• High: syncCoverage() + materialize() added at top of WarpIntakeAdapter and WarpDashboardAdapter operations.
• High: Cached graphPromise cleared on rejection in both WARP adapters.
• High: ESLint + lint script now covers .tsx files.
• High: AllNodesView quest detail modal now uses correct flatQuests array (was indexing into snapshot.quests).
• High: Dashboard.tsx refresh wrapped in useCallback — removed eslint-disable comment.
• High: RoadmapView navigableIndices moved into useEffect body with proper deps — removed eslint-disable.
• High: xyph-dashboard.tsx logo index fallback ?? 3.
• Medium: Quest.toProps() method; CoordinatorService Phase 3 uses it instead of ...q spread.
• Medium: RebalanceService is now a required constructor param (no hidden default).
• Medium: WarpDashboardAdapter skips scrolls with empty questId; caches getNodeProps across passes.
• Medium: InboxView — modal state captures questId at open time; detailHeight clamped; empty-inbox guard on p/x; arrow keys guarded in rationale modal; error state stored as parsed code+message.
• Medium: Shared status-colors.ts module — all views import from one source (includes INBOX).
• Medium: Typed QuestNode.status as QuestStatus, added CampaignStatus and ApprovalGateStatus types.
• Medium: graveyard-ghosts.mts skips patch.commit() when no mutations.
• Low/Nit: ?? instead of || for env reads; positive agentId regex; asciiBar fill clamped; Scrollbar thumbStart clamped; PageUp/PageDown in all scrollable views; trailing newlines on logo files; docs fixes (CHANGELOG formatting, README spacing, RFC milestone/author, model name).
• Tests: getOutgoingEdges added to mocks; service construction in beforeEach; test names corrected; mock extraction type-narrowed; ordering dependency documented.

Wave 8 — Final review fixes (CR-005)

• Critical: Dashboard graph cache invalidation — WarpDashboardAdapter.invalidateCache() clears cached graph via WarpGraphHolder.reset(), called on every refresh so intake mutations (promote/reject) are visible immediately. Added invalidateCache?() to DashboardPort.
• Major: GuildSealService.generateKeypair() now uses loadKeyring() for validation parity with verify(), eliminating divergent schema checks. Orphaned .sk file rollback on keyring write failure prevents permanently broken state. verify() now catches loadKeyring() errors gracefully (returns false instead of throwing).
• Major: LandingView progress bar now excludes GRAVEYARD quests (was only excluding INBOX), consistent with milestone detection logic.
• Major: ESLint test block now references tsconfig.test.json (was tsconfig.json which doesn't include test/**).
• Minor: WarpIntakeAdapter validates task: prefix on questId in all three methods (promote/reject/reopen). WarpRoadmapAdapter.getOutgoingEdges() now calls syncCoverage()/materialize() before reading edges. InboxView status bar clarifies only promote requires human.*. Static randomBytes import in GuildSealService.
• Fix: Restored ajv, ajv-formats, ajv-errors to package.json — incorrectly removed in CR-003 (L-03/L-04) but still required by validatePatchOps.ts. CI verify-patch-ops now passes.
• Doc: Updated CLAUDE-XYPH-PAUSE.md — DSH-004 marked resolved; DSH-001 remains sole pre-merge blocker. CHANGELOG version header updated to 1.0.0-alpha.5.

Wave 7 — Continued review fixes (CR-004)

• Medium: Extracted shared WarpGraphHolder helper — eliminated triplicated getGraph()/initGraph() boilerplate across Dashboard, Intake, and Roadmap adapters (M-25).
• Low: Added runtime isNeighborEntry / toNeighborEntries type guard — replaced unsafe as NeighborEntry[] casts in all adapters with validated filtering (L-20). 10 new unit tests.
• Nit: QuestDetailPanel no longer receives full GraphSnapshot — callers pre-resolve campaign/intent titles (N-16).
• Low: Added 2-line scroll margin to moveSelection in all 4 scrollable views — selection no longer sits at the very edge of the visible area (L-30).
• Documented design decisions: campaign-aware rebalancing deferred to Milestone 6 (M-13); isHumanPrincipal convention-based, not security boundary (M-19); isActive input-focus architecture documented (M-26); view remount on tab switch acknowledged as Ink limitation (M-27); campaign:TRIAGE → graveyarded roadmap:ROOT edge documented (L-08); agent-only requestedBy per Constitution Art. IV.2 (L-14); instant resolution intentionally allowed (L-15); trustDir CWD dependency documented in JSDoc (L-18); private key memory handling is a JS limitation (L-19); upsertQuest can't unset properties, needs tombstone convention (L-24); hint text location clarified (N-21). N-02, N-03, N-04 were already addressed in CR-002.
• All 113 code review issues resolved; CODE-REVIEW-ISSUES.md tracking file removed. 186 tests passing.

Wave 6 — 26 additional issues resolved (CR-003)

• High: Documented order-independent integration test design (H-10); each test now uses dedicated seed nodes.
• Medium: Removed chalk from domain-layer CoordinatorService (M-12); validated task: prefix on quest CLI command (M-43); verified first quest upserted in partial-failure test (M-39); added modal staleness guard in InboxView (M-34); nearest-neighbor selection fallback in LineageView (M-36).
• Low: ESLint now covers root-level entry points (L-02); removed dead dependencies ts-node, ajv, ajv-errors, ajv-formats (L-03, L-04); explicit type narrowing in graveyard-ghosts.mts (L-06); documented asymmetric heartbeat behavior (L-11); graceful daemon shutdown with 500ms drain (L-12); filterSnapshot now removes scrolls referencing GRAVEYARD quests (L-17); documented zero-hour quests as intentional (L-16); negative hours clamped to 0 in WarpRoadmapAdapter (L-25); promote() verifies intentId/campaignId exist before creating edges (L-22); error state bypasses landing screen (L-27).
• Nit: Documented --campaign "none" escape hatch (N-24); clarified noUncheckedIndexedAccess guard comment (N-26).
• 176 tests passing (was 172 pre-review).

NIT Remediation — 27 items resolved (CR-002)

• Config: Removed redundant @typescript-eslint/eslint-plugin and @typescript-eslint/parser devDependencies (bundled by typescript-eslint v8+). Removed 8 redundant strict sub-options from tsconfig.json (implied by "strict": true).
• Type safety: ApprovalNode.trigger typed as ApprovalGateTrigger (was string); STATUS_COLOR typed as Record<string, StatusColor> — removed all as StatusColor casts. WarpDashboardAdapter now validates trigger values.
• GuildSealService: Renamed canonicalPayload → serializePayload; replaced scroll as unknown as Json double-cast with explicit field mapping.
• render-status.ts: Merged dual scroll maps into single iteration; fixed orphan tree connectors (├─/└─).
• TUI views: Spacer row keys use absolute index; removed redundant scroll-clamping useEffect in all 4 views; LandingView nextUp refactored to .slice(1).map(); fixed campaign selection removing empty-campaign false positive; InboxView agentId truncated in error; onMutationEnd() called before setModal(null).
• Entry points: Removed dead truthiness check on required --campaign; renamed __filename/__dirname to currentFilePath/currentDir; added noUncheckedIndexedAccess comment on logo fallback.
• Docs: CHANGELOG comparison links; README Milestone 6 cross-reference and Omega footnote; RFC_001 milestone: → campaign: taxonomy fix and Section 8 clarification; removed model attribution from pause notes.
• Tests: Mock addEdge returns Promise<string>; DashboardService test uses direct snapshot access instead of mock internals; added ApprovalGate boundary tests (instantaneous resolution, negative createdAt); removed duplicate IntakeService test.

1.0.0-alpha.4 - 2026-02-17

Added — POWERLEVEL™ Refactor: Genealogy of Intent Activation

Full Orchestration Pipeline (PL-001)

• Integrated TriageService and RebalanceService into the CoordinatorService heartbeat and orchestration flow.
• The orchestrate method now accepts an optional contextHash (BLAKE3) to link mutations back to their originating human intent.
• Every quest produced in a batch now passes through the RebalanceService for campaign-level resource constraint checking (160h limit).

Structured Ingestion (PL-002)

• Refactored IngestService to defer entity validation to the NormalizeService phase.
• Replaced "Silent Failure" (regex continue on invalid fields) with a permissive factory that allows the validation layer to provide structured error feedback.
• Bypassed Quest constructor validation during raw ingestion using Object.create(Quest.prototype) to allow the full pipeline to report all violations at once.

Verification (PL-003)

• Added test/unit/CoordinatorService.POWERLEVEL.test.ts: A high-fidelity test suite covering the Golden Path (Intent Linking), failure modes (Rebalance/Validation), and stress scenarios (Swarm orchestration).

1.0.0-alpha.3 - 2026-02-17

Added — Milestone 4: SOVEREIGNTY

Intent node type & actuator command (SOV-001)

• intent: prefix added to schema PREFIXES.
• authorized-by edge type added to schema EDGE_TYPES.
• New Intent domain entity: enforces intent: ID prefix, title ≥ 5 chars, and requestedBy must start with human. (agents cannot be sovereign roots — Constitution Art. IV).
• xyph-actuator intent <id> --title "..." --requested-by human.<name> command declares a human-signed sovereign Intent node in the graph.
• xyph-actuator quest gains optional --intent <id> flag to create an authorized-by edge linking a Quest to its Intent root.

Constitutional enforcement — Genealogy of Intent (SOV-002)

• RoadmapPort gains getOutgoingEdges(nodeId), implemented in WarpRoadmapAdapter via graph.neighbors(id, 'outgoing').
• New SovereigntyService: checkQuestAncestry(questId) validates the presence of an authorized-by edge to an intent: node; auditBacklog() scans all BACKLOG quests and returns violations.
• xyph-actuator quest now hard-rejects if --intent is absent (Constitution Art. IV — Genealogy of Intent). Exit 1 with a constitutional violation message.
• New xyph-actuator audit-sovereignty command: scans all BACKLOG quests and reports violations with remediation hint.

Approval gate node type (SOV-003)

• approval: prefix added to schema PREFIXES.
• approves edge type added to schema EDGE_TYPES.
• New ApprovalGate entity: enforces approval: ID prefix, requestedBy must be an agent (agent.), approver must be a human (human.), resolvedAt ≥ createdAt when present, trigger must be CRITICAL_PATH_CHANGE or SCOPE_INCREASE_GT_5PCT (Constitution Art. IV.2).

Guild Seal cryptographic signing on scrolls (SOV-004)

• New GuildSealService: Ed25519 keypair generation, detached signing, and verification of Scroll payloads.
• Canonical payload: { artifactHash, questId, rationale, sealedBy, sealedAt } hashed with blake3.
• Private keys stored in trust/<agentId>.sk (gitignored, mode 0600); public keys registered in trust/keyring.json.
• xyph-actuator seal now attaches guild_seal_alg, guild_seal_key_id, guild_seal_sig, sealed_by, sealed_at, and payload_digest to every Scroll node. Warns (non-fatal) if no private key found.
• New xyph-actuator generate-key command generates and registers a keypair for the active agent.
• Keypair generated and registered for agent.james (did:key:agent.james).

Other

• trust/*.sk added to .gitignore.
• Declared intent:SOVEREIGNTY as the root Intent for the entire project.
• 30 new unit tests (Intent, SovereigntyService, ApprovalGate, GuildSealService). Total: 101 passing.

Breaking Changes

• xyph-actuator quest now requires --intent <id>. Any quest creation without a sovereign Intent root is rejected with exit code 1.

1.0.0-alpha.2 - 2026-02-15

Changed

• Renamed Task entity to Quest with QuestStatus, QuestType, QuestProps (Digital Guild terminology).
• Quest constructor now enforces invariants: task: prefix, title >= 5 chars, finite non-negative hours.
• RoadmapPort methods renamed: getTasks→getQuests, getTask→getQuest, upsertTask→upsertQuest.
• RoadmapPort.addEdge now uses EdgeType union instead of bare string.
• Added tsconfig.test.json for test-file type-checking.
• declarations.d.ts: executeStream return type now matches git-warp's expected StreamResult interface.

Fixed

• Merged duplicate ### Added sections in CHANGELOG.
• Dockerfile: non-root user, --no-install-recommends, accurate CMD comment.
• ESLint: added ignores for dist/** build artifacts.
• Quoted lint glob in package.json to prevent shell expansion.
• declarations.d.ts: fixed invalid static in interface, replaced any with unknown.
• coordinator-daemon.ts: validated INTERVAL_MS, added graceful shutdown (SIGINT/SIGTERM), circuit breaker on repeated failures.
• IngestService: removed unused TaskStatus import, derived checkbox status from regex capture group instead of line.includes('[x]').
• NormalizeService: normalize() returns array directly instead of no-op .map().
• CoordinatorService: dependency injection via constructor, per-task error handling with result accumulation, proper Task import instead of inline import() type.
• WarpRoadmapAdapter: extracted buildTaskFromProps with safe runtime type checks, fixed falsy checks on claimedAt/completedAt timestamps.
• inspect-graph.ts: changed forEach callback to block body to avoid implicit return.
• schema.ts: validate before casting in validateNodeId and validateEdgeType.
• xyph-actuator.ts: centralized createPatch helper, normalized default agent ID constant, added --hours validation via InvalidArgumentError.
• Tests: beforeEach mock reset to prevent state leaks, vi.mocked() instead of brittle double-cast, added isClaimed() tests, expanded NormalizeService test coverage, renamed misleading test description.

1.0.0-alpha.1 - 2026-02-15

Added

• Rebalance Service: Implemented RebalanceService to enforce the 160-hour limit per campaign (Phase 4).
• CodeRabbit Integration: Added .coderabbit.yaml to enable AI code reviews on all branches, facilitating stacked PR workflows.
• Triage Service: Introduced TriageService for backlog auditing and linking tasks to human intent (origin_context).
• Task Entity Expansion: Formalized originContext property in the domain model and persistence layer.
• Orchestration FSM: Integrated Ingest and Normalize services into CoordinatorService to provide a full pipeline from raw input to graph mutation.
• Canonical Corpus: Initialized docs/canonical/ with 15 foundational specifications (Constitution, Agent Charter, Orchestration Spec, etc.).
• XYPH Actuator: Implemented xyph-actuator.mjs for Quest management (Initialize, Claim, Seal) using the git-warp Node.js API.
• Graph Schema: Formalized node/edge taxonomy and runtime validators in src/schema.js.
• Roadmap Bootstrap: Initialized xyph-roadmap with Milestone 1 (BEDROCK) tasks.
• Squadron Integration: Formalized Digital Guild principles within the agentic workflow.
• Infrastructure Patch: Applied @mapbox/node-pre-gyp patch to resolve DEP0169 warnings in modern Node.js.
• Inspection Tooling: Created src/inspect-graph.js for deep graph state analysis.

Changed

• TS Execution Refinement: Switched from ts-node to tsx for CLI execution to resolve DEP0180 (fs.Stats) deprecation warnings in Node 22+.
• Setup Script Improvements: Enhanced scripts/setup-milestone-2.js with idempotency checks and robust error propagation (non-zero exit codes).
• Actuator Refinement: Added ESM shebang to xyph-actuator.ts for direct execution.
• Normalize Service: Implemented NormalizeService for task enrichment and constitutional validation (Phase 2).
• Ingest Service: Implemented IngestService for parsing Markdown-based task definitions into domain entities (Phase 1 of Orchestration Pipeline).
• TypeScript Migration: Full project conversion to strict TypeScript with zero any tolerance.
• Hexagonal Architecture: Established clean boundaries with RoadmapPort and WarpRoadmapAdapter.
• Coordinator Daemon: Initial implementation of the CoordinatorService and heartbeat loop.
• Dockerized Testing: Integrated Vitest with a node:22-slim Docker environment for isolated verification.
• Strict Linting: Configured ESLint with typescript-eslint strict rules.
• Refined Actuator syncWith logic to use syncCoverage() for reliable multi-writer convergence.