The plugin directory Claude Code loads. Single source of truth for the framework.

forgebee/
├── .claude-plugin/
│   └── plugin.json              # Plugin metadata (name, version, hooks)
├── INDEX.md                     # Auto-generated routing index — loaded on SessionStart
├── hooks/
│   ├── hooks.json               # Hook event wiring
│   └── scripts/                 # 25 wired lifecycle hooks (+ _common.js shared helper)
├── agents/
│   ├── *.md                     # 45 specialist agent personas
│   └── references/              # Reference material extracted from agents (W16 bloat trim)
├── commands/                    # 38 slash commands
├── contexts/                    # Session modes (dev, research, review)
├── rules/                       # Language-specific conventions (common, TS, PHP, Python)
├── skills/
│   ├── audit-self/              # On-demand quality scorecard with regression detection
│   ├── brainstorming/           # Opt-in hard-gate via /workflow --strict
│   ├── checkpoint-preview/      # Diff-by-concern review before Code Debate
│   ├── code-{advocate,skeptic,judge}/      # Code debate triad (context: fork)
│   ├── continuous-learning/     # Instinct-based learning system + references/
│   ├── elicitation/             # 18 reasoning methods (methods.csv)
│   ├── figma-code-sync/         # Code-first Figma design-system reconciliation
│   ├── forgebee-setup/          # Project initialization
│   ├── investigate/             # Forensic case files (Confirmed/Deduced/Hypothesized)
│   ├── project-router/          # Stack detection + routing
│   ├── requirements-{advocate,skeptic,judge}/ # Requirements debate triad (context: fork)
│   ├── review-*/                # 11 focused review skills (context: fork)
│   ├── review-all/              # Inline pre-push quality gate
│   ├── strategy-{advocate,skeptic,judge}/  # Strategy debate triad (context: fork)
│   ├── surface-ambiguity/       # Catch silent picks mid-task
│   └── terse-report/            # Sub-agent token compression for orchestrators
├── templates/                   # Decision log, addendum, failure-capture, investigation case file, prompt-defense baseline
└── eval/                        # Eval scenarios

• P1 Trace test — every changed line traces to the user's request; no drive-by edits
• P2 Senior engineer test — "would a senior engineer call this overcomplicated?" before reporting DONE
• P3 YAGNI timing — solve today's problem simply, not tomorrow's prematurely
• P4 Orphan rule — clean only what your changes made unused
• P5 Anti-stop rule — orchestrators continue with next-step work immediately after dispatch
• P6 Severity standard — Critical / High / Medium / Low across all review skills

• Two-stage review in /workflow: Spec Compliance Check → Checkpoint Preview → Code Debate → Deliver
• Adversarial debate triads (advocate/skeptic/judge) for requirements, code, and strategy
• 3-failed-fix Iron Law in debugger-detective — escalate to architecture after 3 attempts
• Failure-Capture template (7 fields) required before any recovery action
• Verification-enforcer + delivery-agent contract — Step 0 reads verdict, does NOT re-run

• Routing index (INDEX.md) — Claude reads 1 file once per session vs. scanning every agent/skill/command frontmatter block
• Terse-report mode — sub-agent reports compress ~65% via responseStyle: "orchestrator" contract
• Learnings compression — /learn ages entries >14 days, archives originals
• Bloat-trimmed agents — 6 worst offenders moved to references/ (1,733 lines extracted)

• Adversarial Input Hardening preamble on all 45 agents (homoglyphs, urgency, role-play overrides flagged)
• Budget circuit breaker on every Task() dispatch (maxHops default 8, ceiling 64) with constant-string errors
• Defensive hooks — safeWriteFlag (O_NOFOLLOW symlink defense), validateHookFields (settings.json guard)
• Sensitive-path refusal in compression (.env, credentials, .ssh/, .aws/, private keys)

• 38 slash commands — orchestration, diagnosis, quality, growth, learning, meta
• 45 specialist agents — code, growth, debate, WordPress, Next.js, mobile, CRO, tool
• 34 skills — 13 inline + 21 context:fork
• 25 lifecycle hooks across 10 events
• 6 templates — decision log, addendum, failure-capture, investigation case file, prompt-defense baseline, brainstorming/spec

• Multi-platform — works with Claude Code, Codex, Cursor, Gemini, OpenClaw (separate manifests)
• Adaptive pipeline — /workflow scrum phase is promptable (full sprint planning OR direct delegation)
• Decision logs + addenda — /workflow and /plan persist decisions across runs
• Continuous learning — /learn extracts instincts; /evolve clusters them; auto-nudge on SessionStart
• /audit-self — re-runs the quality scorecard, surfaces regressions since last audit
• /elicit — 18 named reasoning methods (Pre-mortem, Red Team, Inversion, Stakeholder Round Table) applied to artifacts
• Growth OS — 9-phase marketing pipeline with strategy debate and CRO
• Project management — state.yaml, TASKS.md, automated dashboards
• Instruction priority — CLAUDE.md > Inline skills > Forked skills > Subagents > Defaults

/plugin marketplace add forbee-dev/ForgeBee
/plugin install forgebee@forbee-dev

For full documentation, see the main README. For the routing index, see INDEX.md — auto-generated, never hand-edit.