{
  "markdown": "# Agent Triforce\n\n[![GitHub stars](https://img.shields.io/github/stars/ArtemioPadilla/agent-triforce?style=flat-square)](https://github.com/ArtemioPadilla/agent-triforce/stargazers)\n[![License: MIT](https://img.shields.io/badge/License-MIT-yellow?style=flat-square)](LICENSE)\n[![Version](https://img.shields.io/badge/version-0.7.0-blue?style=flat-square)](CHANGELOG.md)\n[![Claude Code](https://img.shields.io/badge/Claude_Code-Plugin-blueviolet?style=flat-square)](https://claude.ai/code)\n[![Sponsor](https://img.shields.io/badge/sponsor-EA4AAA?style=flat-square&logo=github-sponsors&logoColor=white)](https://github.com/sponsors/ArtemioPadilla)\n\nA 3-agent development system (PM / Dev / QA) that applies checklist methodology from *The Checklist Manifesto* (Atul Gawande) and Boeing's checklist engineering (Daniel Boorman) to AI-assisted software development. Built for teams and individuals who want structured, auditable, production-grade workflows — not ad-hoc prompting.\n\n![Agent Triforce Dashboard](docs/images/dashboard.png)\n\n## Why Agent Triforce?\n\nMost multi-agent systems focus on *more agents*. Agent Triforce focuses on *more discipline*.\n\n- **Methodology over quantity** — 3 agents with 24 checklists (117 items) governed by the WHO Surgical Safety Checklist model, not 100+ generic agents with no coordination\n- **Three mandatory pause points** — every agent invocation follows SIGN IN (before work), TIME OUT (mid-workflow verification), SIGN OUT (before handoff). Steps don't get skipped.\n- **Engineering principles built in** — Clean Architecture, TDD, FIRST testing, and refactoring methodology are embedded into agent behavior, not left to chance\n- **Persistent memory** — agents build on prior decisions across sessions. Context carries forward, not lost between invocations\n- **Structured handoffs** — 6 defined communication paths between agents with explicit protocols (what was done, what to watch for, what's needed next)\n\n## Install\n\n### As a Plugin (add to any existing project)\n\n```bash\n/plugin marketplace add ArtemioPadilla/agent-triforce\n/plugin install agent-triforce@agent-triforce\n/agent-triforce:setup    # scaffold project directories\n```\n\nFor a step-by-step walkthrough, see the [Quickstart Guide](docs/quickstart.md).\n\n### As a Template (start a new project)\n\n```bash\ngh repo create my-project --template ArtemioPadilla/agent-triforce\ncd my-project\nclaude\n```\n\n## Agents\n\n| Agent | Name | Role | Model | Mode |\n|-------|------|------|-------|------|\n| PM | **Prometeo** | Product specs, business logic, prioritization | Sonnet | Plan (read-only) |\n| Dev | **Forja** | Architecture, implementation, testing, docs | Inherit | Accept Edits |\n| QA | **Centinela** | Security, code review, compliance, dead code | Sonnet | Default |\n\n## Skills\n\n| Skill | Agent | What it does |\n|-------|-------|-------------|\n| `/feature-spec` | Prometeo | Create a feature specification (tiered: S/M/L) |\n| `/implement-feature` | Forja | Implement a feature from its spec (TDD workflow) |\n| `/generate-tests` | Forja | Generate tests following FIRST + Arrange-Act-Assert |\n| `/review-findings` | Forja | Fix findings from a QA review |\n| `/reindex` | Forja | Rebuild codebase knowledge index |\n| `/security-audit` | Centinela | Deep OWASP security audit |\n| `/code-health` | Centinela | Dead code, tech debt, dependency scan |\n| `/release-check` | Centinela | Pre-release verification gate with GO/NO-GO |\n| `/traceability` | Centinela | Spec-to-implementation traceability matrix |\n| `/checklist-health` | Centinela | Checklist evolution tracking and health report |\n| `/simulate-failure` | Centinela | Non-Normal procedure training (fire drills) |\n\nWhen installed as a plugin, skills are namespaced: `/agent-triforce:feature-spec`, etc.\n\n## Workflow\n\n```\n/feature-spec → /implement-feature → /security-audit → /review-findings → repeat\n```\n\n```\nPM  SIGN IN → spec → TIME OUT → SIGN OUT\n  → Dev SIGN IN → implement → TIME OUT → TIME OUT → SIGN OUT\n    → QA  SIGN IN → audit → TIME OUT → SIGN OUT\n      → Dev SIGN IN → fix → SIGN OUT\n        → QA  SIGN IN → verify → SIGN OUT\n```\n\n## Methodology\n\nThis system applies principles from *The Checklist Manifesto* (Atul Gawande) and Boeing's checklist engineering (Daniel Boorman), structured around the WHO Surgical Safety Checklist.\n\n**Core ideas:**\n- **Checklists supplement expertise** — reminders of the most critical steps, not how-to guides\n- **FLY THE AIRPLANE** — step 1 of any emergency is to remember your primary mission\n- **Three pause points** per agent invocation: **SIGN IN** (before work), **TIME OUT** (mid-workflow verification), **SIGN OUT** (before handoff)\n- **Two types**: DO-CONFIRM (verify after work) and READ-DO (step-by-step for handoffs and error recovery)\n- **24 checklists, 117 items** across 3 agents. No checklist exceeds 9 items.\n\n## Engineering Principles\n\nEvery agent follows established software engineering practices:\n\n- **Specs**: IEEE 830 + INVEST validation, tiered templates (S/M/L), GIVEN/WHEN/THEN acceptance criteria\n- **Architecture**: Clean Architecture (Robert C. Martin) — dependency rule, screaming architecture, SOLID\n- **Development**: TDD Red-Green-Refactor (Kent Beck), Clean Code (<30 line functions, meaningful names, DRY)\n- **Testing**: FIRST principles (Fast, Isolated, Repeatable, Self-validating, Timely), Arrange-Act-Assert pattern\n- **Refactoring**: Fowler's toolkit — Extract Method, Rename, Move, Boy Scout Rule\n- **Infrastructure**: 12-Factor App patterns, Gang of Four design patterns\n\n## Key Features\n\n- **Persistent Memory**: Each agent remembers decisions across sessions, with cross-agent conflict detection\n- **Auto-formatting**: Dev agent auto-runs ruff/biome after every file edit\n- **Permission Enforcement**: PM can't edit code (plan mode), QA is read-only by default\n- **Security Scanner**: Pre-commit hook detects hardcoded secrets, SQL injection, XSS patterns\n- **System Dashboard**: HTML dashboard auto-generated after every agent session\n- **Automated Handoffs**: Structured handoff artifacts generated at every agent boundary\n- **Approval Gates**: Formal plan and release gates between agents with audit trail\n- **Workflow Tracking**: `/status` command shows current phase, checklists, blockers\n- **Session Analytics**: Per-agent token usage, cost estimates, and performance tracking\n- **CI/CD Templates**: GitHub Actions workflows for PR review, security audit, release check\n- **Codebase Index**: Shared module map with dependency graph, hotspots, and signatures\n- **Smart Routing**: Per-skill model selection (Haiku/Sonnet/Opus) with cost estimates\n- **Growth Tracking**: Pre-launch readiness gates, milestone-gated actions, weekly metrics log\n\n## Project Structure\n\n```\n.\n├── CLAUDE.md                          # System rules & conventions\n├── CHANGELOG.md                       # Keep a Changelog format\n├── TECH_DEBT.md                       # Technical debt register\n├── .claude/\n│   ├── agents/\n│   │   ├── prometeo-pm.md             # PM agent\n│   │   ├── forja-dev.md               # Dev agent\n│   │   └── centinela-qa.md            # QA agent\n│   └── skills/                        # 11 skill definitions\n├── agent-triforce/                    # Plugin marketplace package\n│   ├── agents/                        # Agent configurations\n│   ├── skills/                        # Skill definitions\n│   ├── commands/                      # 15 commands\n│   ├── hooks/                         # security scanner, handoffs, dashboard\n│   └── tools/                         # Tool scripts\n├── tools/\n│   ├── dashboard.py                   # HTML/terminal system dashboard\n│   ├── security-scanner.py            # Pre-commit secret/vulnerability scanner\n│   ├── handoff-generator.py           # Structured handoff artifact generator\n│   ├── workflow-tracker.py            # Workflow state and progress tracker\n│   ├── gate-checker.py                # Plan/release approval gates\n│   ├── memory-sync.py                 # Cross-agent memory conflict detection\n│   ├── traceability.py                # Spec-to-implementation traceability\n│   ├── session-tracker.py             # Session analytics and cost tracking\n│   ├── growth-tracker.py              # Growth phase tracking and readiness gates\n│   └── codebase-indexer.py            # Codebase knowledge index builder\n├── templates/\n│   ├── lsp/                           # LSP configs (Python, TypeScript, Rust)\n│   ├── mcp/                           # MCP configs (SonarQube, Linear, Jira)\n│   ├── ci/                            # GitHub Actions workflow templates\n│   ├── agent-routing.json             # Per-skill model routing\n│   └── context-profiles.json          # Context loading profiles\n├── docs/\n│   ├── specs/                         # Feature specifications (PM)\n│   ├── adr/                           # Architecture Decision Records (Dev)\n│   ├── reviews/                       # Code reviews & audits (QA)\n│   ├── handoffs/                      # Handoff artifacts between agents\n│   ├── gates/                         # Approval gate documents\n│   ├── traceability/                  # Spec-to-code matrices\n│   └── training/                      # Simulation training reports\n├── src/                               # Source code\n└── tests/                             # Tests\n```\n\n## Customization\n\n| What | Where |\n|------|-------|\n| Tech stack preferences | `CLAUDE.md` → Tech Stack Preferences |\n| Agent behavior | `.claude/agents/{agent-name}.md` |\n| Add new skills | `.claude/skills/{skill-name}/SKILL.md` |\n| Project-specific rules | `CLAUDE.md` or `.claude/rules/` |\n\n## Requirements\n\n- [Claude Code](https://claude.ai/code) v2.1.32+\n- Claude Max or Pro subscription (or API key)\n\n> **Note**: Agent Triforce is a provider-agnostic methodology. The checklist framework, agent roles, and workflows are portable to other AI coding tools as they mature to support multi-agent orchestration.\n\n## Support\n\nIf Agent Triforce helps your team ship better software, consider [sponsoring development](https://github.com/sponsors/ArtemioPadilla).\n\n## License\n\n[MIT](LICENSE)\n",
  "bytes": 10084,
  "sha": "ab2cea61c0df6d74edbe7c25b5128961f6e413eae8e46890f5493afe5c93bd20",
  "repo_slug": "artemiopadilla/agent-triforce",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_artemiopadilla_agent_triforce_63127d8b/readme"
}