Skip to content

CHANGELOG.md

Latest commit

 

History

History
285 lines (231 loc) · 19.3 KB

CHANGELOG.md

File metadata and controls

285 lines (231 loc) · 19.3 KB

Changelog

All notable changes to BA-Kit Antigravity.

[1.5.0] - 2026-04-14 — E2E Quality Gate + Mini-App Chấm Công Showcase + Anti-Rationalization Reconciliation (Marketing v3.5.0)

Reconciliation note: This release also merges the parallel v3.2.0 + v3.3.0 anti-rationalization work that shipped on the main branch while this feature branch was developing the Sprint Spine. Both tracks are now unified. Agent count: 44 (33 original with anti-rationalization pattern + 11 new sprint spine agents). The 11 new sprint spine agents inherit the anti-rationalization convention as follow-up work.

Added

  • Phase 08 — CI-integrated E2E quality gate:
    • .agent/scripts/ba_e2e_test.py (1,389 LOC) — 5-layer orchestrator runner (L1 script smoke → L2 helper smoke → L3 skill frontmatter/sections → L4 deep BRD↔US↔API↔test consistency → L5 fixture coverage), --report markdown output, --json stable schema v1.0.0 (ensure_ascii=False for Vietnamese round-trip), --fixture selector
    • .github/workflows/e2e-skills.yml — GitHub Actions quality gate triggered on PR + push to main/feat/** touching .agent/, docs/, outputs/, or the workflow itself. Writes markdown to GITHUB_STEP_SUMMARY, uploads reports/ artifact (14-day retention), posts sticky PR comment via marocchino/sticky-pull-request-comment@v2. Fails only on FAIL/CRASH verdict (warnings don't block). Python 3.11, stdlib only, 10-min timeout
    • README badge: GitHub Actions status linked to the e2e-skills workflow
    • L4 deep lints — single-pass FixtureInventory + 6 lint families (duplicate US IDs, BRD→US cross-refs, API-spec US refs, test-case US refs, RTM US refs, orphaned test cases). 34 checks on mini-app-cham-cong fixture
    • L2 helper smoke — 14 test cases covering ba_as_built, ba_retro, ba_learn, ba_baseline, ba_second_opinion, ba_setup stdlib interfaces
  • Mini-App Chấm Công — comprehensive BA dogfooding showcase (outputs/mini-app-cham-cong/):
    • 4 BRDs (Nhân viên, Quản lý, HR-Admin, IT/System-Admin) + Demo Plan Sprint-8 + modules overview + Confluence manifest
    • 12 modules (M01-M12) with README + api-spec + db-schema + test-cases + US files each
    • 47 US files across M01-M11 (chấm công, OT, nghỉ phép, công tác, phân ca, chính sách, mobile, báo cáo) rewritten R3 with concrete Gherkin scenarios sourced from Edge Cases tables
    • 6 US files for M12 (quản trị hệ thống — chi nhánh, audit log, offboarding, chốt công, onboarding, data retention)
    • EAMS v2.1 (Employee Attendance Management Specification) with §15.3/§17.5/§17.6 expanded
    • RTM (Requirements Traceability Matrix) full cross-ref
    • AUDIT-REPORT R3 + INDEX + 111/111 Confluence sync complete (0 rendering errors)
  • ba-wiki canonical sections — added ## Input Validation + ## System Instructions to the operation-based template so L3 frontmatter lint passes (6/8 → 8/8)

Changed

  • coverage_checker.py — recognizes lowercase us-* filenames + test-cases.md (case-insensitive US ID regex, dual filename patterns)
  • README badges point to branch=main (was feat/mini-app-cham-cong-docs)
  • Logo cache-bust v3.4.0 → v3.5.0 in both READMEs

Fixed

  • Confluence DC rendering blockers: added json/gherkin language mappings + mermaid-macro support to push scripts; fixed broken link in overview/README
  • R3 Gherkin remediation: 47 US files (M01-M11) rewritten from abstract "Given a user, When they submit, Then success" → concrete scenarios with table-driven Edge Cases
  • M12 R3 remediation: 6 US Gherkin rewrite + EAMS v2.1 §15.3/§17.5/§17.6 + API spec fixes (M03/M04/M09/M11) + BRD cross-ref fixes + Audit R3 + RTM update

Migration notes

  • No breaking changes. Existing v3.4.0 installs continue working
  • To enable the new E2E quality gate on forks: copy .github/workflows/e2e-skills.yml + ensure .agent/scripts/ba_e2e_test.py is present
  • Mini-app Chấm Công docs are a dogfooding showcase — reference material, not executable templates

[1.4.0] - 2026-04-13 — Gstack Distillation (Marketing v3.4.0)

Added

  • 10 new agents distilled from garrytan/gstack analysis:
    • @ba-as-built — spec drift detector (reads git diff, finds where code diverged from BRD/SRS/RTM)
    • @ba-autoreview — strict-sequential meta: consistency → quality-gate → traceability → auditor, optional --dual-voice
    • @ba-retro — time-windowed sprint retro (churn, gate rejection trends, cycle time, per-author leaderboard)
    • @ba-learn — per-project JSONL emergent memory (5 types, PII filter, auto-capture API)
    • @ba-checkpoint — session save/resume for long-running BA work
    • @ba-challenger — 5-vector adversarial red team (unstated / incentive / adversarial / scale-break / sunset)
    • @ba-second-opinion — cross-model independent review (Gemini / OpenAI / Ollama / manual, zero-dep urllib)
    • @ba-baseline — sha256-locked CCB baselines with append-only history + supersede semantics
    • @ba-guard — pre-flight change-control checker (off / warn / strict + optional git hook)
    • @ba-shotgun — N-variant generator for stories / AC / priority / emails (preference capture)
  • 5 Python helpers (zero external deps, stdlib only): ba_as_built.py, ba_retro.py, ba_learn.py, ba_baseline.py, ba_second_opinion.py
  • setup.sh one-liner install with auto-host detection (antigravity / claude-code / cowork)
  • docs/sprint-spine.md — unified 7-phase loop (Discover→Elicit→Define→Validate→Prioritize→Publish→Reflect)
  • AGENTS.md §7 — second-opinion provider auto-detection table (priority order: GEMINI > OPENAI > OLLAMA > manual)
  • AGENTS.md §8 — strict CCB workflow (3-step opt-in for regulated teams)

Changed

  • Agent count 33 → 43 across README, README.vi, agent-cheat-sheet, junior-start, antigravity-protocol, ai-foundation, usage-guide, _shared/system-prompt, ba-master, architecture-decisions
  • README badges bumped: Version-3.1.0 → 3.2.0, Agents-33 → 43, logo cache-bust v3.1.0 → v3.2.0
  • ba-master/SKILL.md Decision Matrix: +20 routing entries for 10 new skills, +Sprint Spine Agents registry section
  • ba-shotgun/SKILL.md — added rationale block explaining N=3 default + "vary constraint axis instead of inflating count"

Fixed

  • YAML frontmatter: 23 SKILL.md files had unquoted description: [Agentic] foo causing strict parser failures. All quoted ("[...]") — 45/45 now parse cleanly.
  • docs/sprint-spine.md self-correction: "v3.1 has 33 agents" → "v3.1 had 33, v3.2 has 43"
  • docs/architecture-decisions.md Production Readiness Score: agent count breakdown updated
  • BA-fit reframing (4 v3.4 skills bumped 1.0.0 → 1.1.0): catch where gstack-imported mechanisms leaked engineer vocabulary into BA-facing skills.
    • ba-as-built — primary mode now reads delivered evidence pack (UAT reports, release notes, demo notes); git mode demoted to opt-in advanced for hybrid BA+dev teams
    • ba-retro — primary input now BA-Kit JSONL metric streams + file mtime; git log demoted to tertiary opt-in; output reframed from "commits/authors/sessions" to "gate pass rate / churn / stakeholder responsiveness"
    • ba-baseline — agent now uses natural-language conversation ("which doc / version / signer / rationale"); sha256 hidden entirely from BA output; integrity check runs silently
    • ba-guard — 3 modes (off / warn / strict) described in BA language; alerts say "BRD-HR was edited 2 days after sign-off — was this re-approved?" instead of "DRIFT detected: hash mismatch"; git pre-commit hook stays as opt-in advanced
    • Python helpers UNCHANGED (mechanism layer is correct for both modes; only SKILL.md presentation reframed)
  • BA-fit Phase 2 — spot-check 5 v3.1 skills + 2 wrappers + 2 connectors:
    • ba-jira + ba-confluence Prerequisites sections reframed: "one-time IT setup, not your daily job" with clear separation between IT setup and BA daily workflow
    • jira-connector + confluence-connector infrastructure header added: "loaded by ba-jira / ba-confluence, not for direct BA invocation"
    • 3 spot-checked v3.1 skills clean (ba-test-gen, ba-data, ba-process)
  • BA-fit Phase 3 — spot-check 5 more v3.1 skills + ship @ba-setup:
    • ba-metrics "Fix the build" engineer slang → "Fix the requirements process" (BA language)
    • 4 v3.1 skills clean (ba-solution, ba-innovation, ba-systems, ba-strategy)
    • NEW skill @ba-setup (44 agents total): one-time setup wizard with BA-friendly natural-language flow for Jira / Confluence / second-opinion provider configuration
    • NEW helper ba_setup.py (zero-dep stdlib): credential validation, atomic .env write (chmod 0600), placeholder rejection, masked token display, connectivity test via urllib
    • Closes the last identified BA-fit gap — BA never types .env files or CLI flags
  • Cumulative: 15/46 skills spot-checked (33%), 0 hard issues, 4 minor surface edits applied. Confidence HIGH that remaining 31 are BA-appropriate.

Historical entries below — these v3.x.y entries reflect the parallel main-branch track (Anti-Rationalization Framework v3.2/v3.3 + license/doc work v3.1.x) that was reconciled into v3.5.0 above. They are preserved unchanged for provenance.

v3.3.0 — 2026-04-13

Added — Full Anti-Rationalization Pattern Coverage

Complete rollout of anti-rationalization pattern to all remaining 28 BA agents. All 33 agents now have the 4 anti-slack sections.

28 agents updated (organized by thematic batches):

  • Batch A — Core Workflow (7): ba-identity, ba-nfr, ba-process, ba-prioritization, ba-solution, ba-conflict, ba-export
  • Batch B — Advanced Analysis (7): ba-metrics, ba-root-cause, ba-innovation, ba-strategy, ba-facilitation, ba-systems, ba-agile
  • Batch C — Quality & Modeling (7): ba-test-gen, ba-quality-gate, ba-consistency, ba-auditor, ba-traceability, ba-data, ba-business-rules
  • Batch D — UX/Delivery/Integration (7): ba-ux, ba-communication, ba-change, ba-diagram, ba-jira, ba-confluence, ba-wiki

Each updated SKILL.md now includes:

  • ## When to Use — explicit triggers + exclusions
  • ## Common Rationalizations — ≥4 domain-specific excuses with factual rebuttals
  • ## Red Flags — ≥4 observable signs the skill is being violated
  • ## Verification — ≥4 evidence-based exit criteria with handoff reference

Stats

  • Pattern coverage: 5/33 → 33/33 (100%)
  • New rationalizations: 140+ (28 × 5 average)
  • New red flags: 140+
  • New verification items: 140+
  • All rationalizations pass specificity test (no copy-pasteable generic filler)

Methodology

Delivered via 4 parallel subagents working on 7 files each, with domain seeds packed into prompts (realistic BA anti-patterns per agent). Quality review: automated section count check + manual specificity/observability/evidence checks on sample files.

v3.2.0 — 2026-04-13

Added — Anti-Rationalization Framework

  • docs/skill-anatomy.md — Formal spec for BA-Kit skill format (required sections, design principles)
  • .agent/skills/using-ba-kit/ — Meta skill: intent-to-skill mapping, lifecycle navigation, anti-slack enforcement
  • .claude/commands/ba-*.md — 6 lifecycle slash commands for Claude Code users:
    • /ba-discover — identity → strategy → questioning → elicitation → ux
    • /ba-analyze — process / data / systems / business-rules
    • /ba-specify — writing → nfr → traceability → diagram
    • /ba-validate — validation → quality-gate → consistency → test-gen → auditor
    • /ba-deliver — export → jira → confluence → communication → change
    • /ba-audit — auditor → traceability → consistency → metrics

Enhanced — Anti-Rationalization Pattern Applied to Key Agents

5 flagship agents updated with new sections (pattern from addyosmani/agent-skills):

  • @ba-master, @ba-elicitation, @ba-writing, @ba-validation, @ba-questioning
  • Each gets:
    • When to Use / When NOT to Use — explicit triggers and exclusions
    • Common Rationalizations — table of excuses + factual rebuttals (prevents skill-skipping)
    • Red Flags — observable signs the skill is being violated
    • Verification — exit criteria checklist with evidence requirements

Changed

  • @ba-master routing table: added @using-ba-kit as fallback for ambiguous intent
  • docs/README.md: added Lifecycle Commands section + Skill Anatomy link

Philosophy

"Skills are processes, not prose. Every skill encodes hard-won engineering judgment as step-by-step workflows with anti-slack mechanisms." — pattern inspired by production-grade engineering skills

v3.1.1 — 2026-04-13

Changed

  • License: MIT → CC BY-NC 4.0 (Creative Commons Attribution-NonCommercial 4.0 International)
  • docs/README.md: Rewritten as onboarding navigation guide with ASCII flowchart reading path
  • All 11 docs files: Complete rewrite for consistency, accuracy, and new-user clarity
    • prompt-library.md: 48 prompts across 10 lifecycle phases
    • workflow-cookbook.md: 15 scenarios (consolidated from 23, removed duplicates)
    • quick-start.md, usage-guide.md, junior-start.md, contributing.md: restructured
    • architecture-decisions.md: 3 new ADRs for v3.1 decisions
    • antigravity-protocol.md: 33 agents roster + diagram tool mandate
    • ai-foundation-for-ba.md, ai-tools-guide.md, design-prototype-guide.md: updated counts

Fixed

  • All stale agent counts (19/21/26) eliminated across docs
  • All broken links in docs/README.md resolved
  • Scenario count inconsistency (23→15) fixed in cheat sheet and ai-foundation
  • Prompt count inconsistency (28/33/45→48) fixed across all files

v3.1.0 — 2026-04-12

Added — 7 New Lifecycle Agents

  • @ba-questioning — Paul-Elder Critical Thinking Framework, interview prep, assumption surfacing
  • @ba-communication — Audience-adapted status reports, executive summaries, meeting minutes
  • @ba-ux — Persona, journey mapping, empathy maps, JTBD, UX psychology, usability testing
  • @ba-data — ERD, data dictionary, DFD, data mapping, migration planning
  • @ba-change — ADKAR assessment, training needs, go-live planning, benefits realization
  • @ba-business-rules — Decision tables, decision trees, rule catalog, conflict detection
  • @ba-diagram — Mermaid v11 (24+ types), BA artifact→diagram mapping, Confluence export

Enhanced

  • @ba-questioning v2.1 — 8 Elements of Reasoning, Intellectual Standards, Wilson Method, Socratic Protocol, bias detection
  • @ba-ux v2.0 — UX Psychology (Cognitive Load, Decision Architecture, Key Laws), heuristic evaluation, usability test protocol
  • @ba-agile — Estimation facilitation (Planning Poker, T-shirt sizing, story splitting)
  • @ba-wiki — Glossary management (ubiquitous language, synonym detection)
  • @ba-confluence — Mermaid diagram handling (Stratus mermaid-macro for DC, pre-render SVG)
  • @ba-master — Routing table expanded with 15+ new entries for all new agents

Knowledge Base

  • 3 new knowledge base files: questioning.md, communication.md, change_management.md
  • 831 entries across 23 domains (unchanged from v3.0)

Documentation

  • 48 copy-paste prompts (was 33)
  • Agent cheat sheet updated for 33 agents
  • Prompt library expanded with 5 new phases
  • .gitignore reorganized: Python artifacts, Claude session state, outputs/, diagrams

Infrastructure

  • outputs/ removed from git tracking (project deliverables are per-instance)
  • .claude/session-state/ removed from git tracking (local per-machine)
  • pycache/*.pyc cleaned from tracking

[1.3.1] - 2026-04-11

Fixed

  • Template paths corrected templates/ → .agent/templates/ across 9 SKILL.md files (21 references) and 7 doc files
  • Stale agent count 32 → 26 corrected in README.md (6), README.vi.md (6), agent-cheat-sheet.md (1)
  • README.vi.md line 149: "32 chuyên gia" → "26 chuyên gia"

[1.3.0] - 2026-04-11

Changed

  • Knowledge entry count corrected 786 → 831 across 12 references in 8 files
  • Template count corrected 13 → 14 across 4 files (prd-template.md was undocumented)
  • .claude-output/CLAUDE.md fully rewritten: 19 → 26 agents, added Integration/Quality/Knowledge sections
  • docs/antigravity-protocol.md updated to v3.0.0, added Quality & Knowledge agent sections
  • README.vi.md repo structure fixed: correct template path, added Knowledge Search section
  • Scenario count standardized to 23 across agent-cheat-sheet, junior-start, ai-foundation
  • Prompt count corrected 28 → 33 in quick-start.md
  • ebooks/README.md version footer updated v2.9.3 → v3.0.0

Fixed

  • Removed 4 phantom AGENT.MD references (file does not exist) in quick-start, contributing
  • docs/ai-foundation-for-ba.md English half said "21 agents" while Vietnamese said "26" — unified to 26
  • docs/junior-start.md typo k# → #, broken path WORKFLOW-COOKBOOK.md → docs/workflow-cookbook.md
  • Logo cache-bust version updated v2.7.0 → v3.0.0 in both READMEs
  • .claude-output/MIGRATION-GUIDE.md expected skill count 25+ → 26+
  • Ebook count corrected 6 → 7 in ai-foundation-for-ba.md

[1.2.0] - 2026-04-11

Added

  • docs/agent-cheat-sheet.md — 26-agent quick reference with workflow chains (was empty)
  • docs/prompt-library.md — 33 copy-paste prompts organized by BA phase (was empty)
  • @ba-wiki added to README, ba-master registry, and agent-cheat-sheet

Changed

  • ba-writing/SKILL.md — standardized AC format (Gherkin + Structured Bullets both accepted)
  • Agent count updated 25 → 26 across all docs (README, README.vi, ba-master, .agent/README)
  • Knowledge entry count corrected 809 → 786 across 6 files

Fixed

  • coverage_checker.py — AC counting regex now matches both #### **ACn. and #### **3.n. formats
  • coverage_checker.py — happy path detection expanded for Vietnamese BA output keywords
  • 9 ambiguous terms replaced with measurable metrics across 7 US output files
  • README.md — broken tree rendering and duplicate brd-template entry
  • Health Score: 78% AT RISK → 97% HEALTHY (post-fix)

[1.1.0] - 2026-04-11

Added

  • ba-traceability SKILL.md — RTM build, impact analysis, health check (was empty)
  • _shared/system-prompt.md — shared identity fragment for all 26 agents
  • .agent/templates/ — 13 BA document templates (BRD, SRS, FRD, US, UC, TC, RTM, etc.)
  • Examples added to 12 skills (P3 + P6)
  • Workflow + Output Format sections added to 13 skills (P3 + P4+P5)

Changed

  • Standardized frontmatter: all 28 skills now have version: 1.0.0
  • docs/README.md — navigation guide with agent activation reference
  • 7 short skills expanded from ~80 → 150+ lines (metrics, conflict, export, root-cause, innovation, solution, identity)
  • 6 C-grade skills expanded with Workflow/Output/Example (nfr, facilitation, process, prioritization, strategy, systems)

Fixed

  • ba-traceability was 0 lines (critical gap) — now 195 lines, Grade A

[1.0.0] - 2026-04-10

Initial Release

  • 26 BA agent skills in .agent/skills/
  • 5 Python scripts (ba_core, ba_search, coverage_checker, gen_docx, batch_remediate)
  • 23 CSV knowledge base files (786 entries across 23 domains)
  • 7 BA ebooks (fundamentals, agile, techniques, career, leadership, systems thinking, requirements memory jogger)
  • 6 documentation guides in docs/
  • Confluence connector + Jira connector skills
  • EAMS Mini App Chấm Công output (first project delivery)