BA-Kit is not a prompt library. It is an agent squad for agentic AI platforms.
It replaces the single-chatbot approach with 44 specialists — each with a defined role, toolset, System 2 Reflective Loop, and anti-rationalization guardrails — running across three platforms:
| Platform | Runtime | Best For |
|---|---|---|
| Antigravity IDE (Google DeepMind) | Agent Skills + MCP | Power users, full toolchain |
| Claude Code (Anthropic) | CLI, Git, CI/CD | Engineers, project-level reasoning |
| Claude CoWork (Anthropic) | Desktop | Non-technical BAs, document synthesis |
User Request
│
▼
@ba-master ──── Triage + Route ──────────────────────────────┐
│ │
▼ ▼
Specialist Agent Context Handoff
(e.g. @ba-writing) (shared continuity)
│
├── System 1: Draft Output
│
├── System 2: STOP & REFLECT
│ ├── Did I hallucinate?
│ └── Verify via grep / python / web
│
└── Verified Output → Handoff to Next Agent
Each agent also ships an anti-rationalization guardrail (v3.2/3.3): explicit "When NOT to Use", common rationalizations with rebuttals, red flags, and verification exit criteria — to prevent skill-skipping under pressure.
Orchestrator (1)
| Agent | Role | Capability |
|---|---|---|
@ba-master |
Dispatcher | Strategy, Routing, Context Management |
Core (3) — The Foundation
| Agent | Role | Capability |
|---|---|---|
@ba-identity |
Chief of Staff | Stakeholder Mapping, RACI, Competency Grid |
@ba-elicitation |
Journalist | Funnel Questioning, Colombo Method |
@ba-writing |
Architect | Visual UI Scan, Gherkin Drafting (BDD) |
Specialized (8) — The Experts
| Agent | Role | Capability |
|---|---|---|
@ba-validation |
QA Lead | Visual QA, Edge Case Detection |
@ba-traceability |
CCB Secretary | Grep Verification, No Hallucinations |
@ba-nfr |
SRE Architect | Web-Validated ISO 25010 Standards |
@ba-process |
Lean Master | Whiteboard Vision, BPMN Waste Analysis |
@ba-prioritization |
Product Manager | MoSCoW, RICE, WSJF Frameworks |
@ba-solution |
Investor | Python-Verified ROI & NPV Math |
@ba-conflict |
Mediator | Harvard Negotiation, ADR Drafting |
@ba-export |
Publisher | Compliance Check, Pandoc Formatting |
Advanced (3)
| Agent | Role | Capability |
|---|---|---|
@ba-metrics |
Data Scientist | SPC Charts, Defect Density, Cpk Stats |
@ba-root-cause |
Investigator | 5 Whys, Fishbone, Pareto Analysis |
@ba-innovation |
R&D Scientist | A/B Testing, Hypothesis Designs |
Strategic (4)
| Agent | Role | Capability |
|---|---|---|
@ba-strategy |
Strategist | PESTLE, SWOT, Business Model Canvas, Porter's 5 Forces |
@ba-facilitation |
Facilitator | Workshop Design, ODEC Framework, Group Dynamics |
@ba-systems |
Systems Analyst | Stocks & Flows, Leverage Points, System Archetypes |
@ba-agile |
Agile Analyst | User Story Mapping, MVP Definition, Hypothesis-Driven |
Quality & Audit (4)
| Agent | Role | Capability |
|---|---|---|
@ba-test-gen |
QA Architect | AC → 7-category Test Cases (BVA, Decision Tables, State Transitions) |
@ba-quality-gate |
Quality Officer | 8-dimension scoring, 5 gates: PASS / CONDITIONAL / REJECT |
@ba-consistency |
Integration Auditor | Cross-artifact alignment: US ↔ API ↔ DB ↔ BRD |
@ba-auditor |
Chief Auditor | Full project health dashboard + action plan |
Lifecycle (7)
| Agent | Role | Capability |
|---|---|---|
@ba-questioning |
Master Interviewer | Context-Free Questions, Interview Prep, Assumption Surfacing |
@ba-communication |
Communicator | Audience Adaptation, Status Reports, Executive Summaries |
@ba-ux |
UX Researcher | Persona, Journey Map, Empathy Map, JTBD, Usability Testing |
@ba-data |
Data Architect | ERD, Data Dictionary, DFD, Data Mapping, Migration Planning |
@ba-change |
Change Agent | ADKAR, Readiness Assessment, Training Plans, Go-Live |
@ba-business-rules |
Rules Engine | Decision Tables, Decision Trees, Rule Catalog, Conflict Detection |
@ba-diagram |
Visualizer | Mermaid v11 (24+ types), BA artifact→diagram, Confluence export |
Integration (2) + Knowledge (1)
| Agent | Role | Capability |
|---|---|---|
@ba-jira |
Jira Bridge | Story→Ticket Transport, Sprint Planning, Transport Gate Reflection |
@ba-confluence |
Confluence Bridge | Markdown→XHTML Publishing, Document Import, Version Tracking |
@ba-wiki |
Knowledge Curator | 2-tier knowledge ingest, wiki query, living docs, glossary management |
| Agent | Phase | Capability |
|---|---|---|
@ba-as-built |
Reflect | Spec drift detector — compares BRD/SRS/RTM against delivered evidence (UAT reports, release notes, demo notes) to find what was built ≠ what was specified. |
@ba-autoreview |
Validate | Strict-sequential meta: consistency → quality-gate → traceability → auditor, with optional --dual-voice. |
@ba-retro |
Reflect | Time-windowed sprint retro: gate pass rate, churn, stakeholder responsiveness — primary input is BA-Kit metric streams (no git required). |
@ba-learn |
Reflect | Per-project JSONL emergent memory (5 types, PII filter, auto-capture API). |
@ba-checkpoint |
any | Session save/resume for long-running BA work (workshops spanning multiple days). |
@ba-challenger |
Validate | 5-vector adversarial red team (unstated / incentive / adversarial / scale-break / sunset). |
@ba-second-opinion |
Validate | Cross-model independent review — Gemini / OpenAI / Ollama / manual (zero-dep urllib). |
@ba-baseline |
Publish | CCB baselines with version + signer + date + rationale. Agent prompts BA in plain language; integrity check happens silently. |
@ba-guard |
any | Pre-flight change-control alert in BA language: "BRD-HR was edited 2 days after sign-off — was this re-approved?" 3 modes (off / warn / strict). |
@ba-shotgun |
Define | N-variant generator for stories / AC / priority / emails (preference capture to @ba-learn). |
📐 Sprint Spine — all 44 agents map into a unified 7-phase loop: Discover → Elicit → Define → Validate → Prioritize → Publish → Reflect. See
docs/sprint-spine.mdfor the full agent → phase mapping.
Antigravity IDE
cp -r ba-kit/.agent/skills/* ~/.gemini/antigravity/skills/Claude Code
cp -r ba-kit/.agent/skills/* ~/.claude/skills/Claude CoWork
cp -r ba-kit/.agent/skills/* ~/Library/Application\ Support/Claude/skills/One-liner install (auto-detects host):
bash .agent/scripts/setup.shThen summon any specialist:
@ba-writing I need a login feature spec.
@ba-solution check the ROI for this initiative.
@ba-master route this request to the right specialist.
Every agent runs a Reflective Cognitive Loop before responding:
- Analyze (System 1) — fast pattern matching
- Draft (System 1) — generate initial output
- Reflect (System 2) — STOP. Did I hallucinate? Verify with
grep/python/ web - Output — verified, high-precision answer
This loop reduces hallucinations and surfaces hidden assumptions automatically.
831 indexed entries across 23 domains, powered by BM25+.
python3 .agent/scripts/ba_search.py "acceptance criteria gherkin"
python3 .agent/scripts/ba_search.py "GDPR compliance" --domain compliance
python3 .agent/scripts/ba_search.py "stakeholder analysis" --multi-domain
python3 .agent/scripts/ba_search.py --statsDomains: writing elicitation validation nfr process prioritization traceability conflict solution systems agile identity workshop innovation metrics modeling ux-research business-rules integration compliance communication testing data-analytics
ba-kit/
├── .agent/skills/ # 44 Agent Skills + 2 Connectors + 1 meta
├── .agent/scripts/ # BM25+ Knowledge Search Engine
├── .agent/data/ # 831 Indexed Knowledge Entries (23 domains)
├── .agent/templates/ # 14 Document Templates (BRD, SRS, FRD, PRD, ...)
├── .agent/wiki/ # Living Knowledge Wiki
├── docs/ # Documentation & Guides
│ ├── agent-cheat-sheet.md
│ ├── prompt-library.md # 48 copy-paste prompts by phase
│ └── knowledge_base/
├── ebooks/ # Synthesized Book Knowledge (9 files)
├── README.md
└── README.vi.md
CC BY-NC 4.0 — Creative Commons Attribution-NonCommercial 4.0 International.
Free for personal, educational, and non-commercial use. Attribution required.
Antigravity • Claude Code • Claude CoWork
Code Less. Think More.