{
  "markdown": "# audit-project\n\nMulti-agent iterative code review that loops until zero issues remain.\n\n## Why\n\nManual code reviews miss things. Single-pass automated reviews find issues but leave fixing to you. audit-project runs up to 10 specialized review agents, fixes the issues it finds, re-reviews the fixes, and repeats until the codebase is clean - or you tell it to stop.\n\nUse it when you want:\n\n- A thorough review before a release or major merge\n- Security-focused auditing of a specific area\n- Quick feedback on recent commits\n- Automated cleanup of accumulated code quality issues\n\n## Installation\n\n```bash\nagentsys install audit-project\n```\n\nRequires [agentsys](https://github.com/agent-sh/agentsys) as the plugin runtime.\n\n## Quick Start\n\n```bash\n/audit-project              # Full iterative review of the entire project\n/audit-project --quick      # Single pass, findings only, no fixes\n/audit-project --recent     # Review last 5 commits only\n```\n\n## How It Works\n\n1. **Context gathering** - Detects project type, framework, languages, and infrastructure (DB, API, CI/CD, frontend/backend).\n2. **Agent selection** - Activates 4 core agents plus up to 6 conditional agents based on what the project contains.\n3. **Multi-agent review** - Each agent reviews only the files relevant to its domain and produces structured findings with file:line locations, severity, and suggested fixes.\n4. **Automated fixes** - Applies fixes in severity order (critical first), batched by file. Rolls back any fix that breaks tests or build.\n5. **Re-review** - Runs agents again on changed files. If new issues appear, fixes and re-reviews again.\n6. **Decision gate** - After each iteration, reports remaining issues and asks whether to continue, create GitHub issues, update TECHNICAL_DEBT.md, or save the queue for later.\n\n### Review Agents\n\n**Always active:**\n\n| Agent | Focus |\n|-------|-------|\n| code-quality-reviewer | Error handling, maintainability, code patterns |\n| security-expert | Vulnerabilities, auth, input validation |\n| performance-engineer | Bottlenecks, algorithms, memory usage |\n| test-quality-guardian | Test coverage gaps, test quality |\n\n**Conditional (activated by project signals):**\n\n| Agent | Activated when |\n|-------|---------------|\n| architecture-reviewer | 50+ tracked files |\n| database-specialist | Sequelize, Prisma, or TypeORM detected |\n| api-designer | Express, Fastify, or NestJS detected |\n| frontend-specialist | .tsx, .jsx, .vue, or .svelte files present |\n| backend-specialist | Server framework detected |\n| devops-reviewer | CI/CD configuration files present |\n\n### Finding Format\n\nEvery finding includes file:line location, severity (critical/high/medium/low), a code quote, a suggested fix, and an effort estimate (small/medium/large).\n\n## Usage\n\n```bash\n/audit-project                      # Full iterative review\n/audit-project src/api              # Review a specific path\n/audit-project --domain security    # Run only the security-expert agent\n/audit-project --quick              # Single pass, no fixes applied\n/audit-project --recent             # Scope to last 5 commits\n/audit-project --create-tech-debt   # Force TECHNICAL_DEBT.md creation\n/audit-project --resume             # Resume from a saved review queue\n```\n\n### Repo Intelligence\n\nWhen [agent-analyzer](https://github.com/agent-sh/agent-analyzer) repo-intel data is available, audit-project uses test-gap signals to prioritize review. Files with high churn but no co-changing test file receive extra scrutiny.\n\n## Requirements\n\n- [agentsys](https://github.com/agent-sh/agentsys) runtime\n- Git repository\n- Works with Claude Code, OpenCode, Codex, and Cursor\n\n## Related Plugins\n\n- [enhance](https://github.com/agent-sh/enhance) - Plugin structure and config analysis\n- [perf](https://github.com/agent-sh/perf) - Dedicated performance investigation with profiling\n- [repo-intel](https://github.com/agent-sh/repo-intel) - Unified static analysis (git history, AST symbols, doc-code sync)\n- [ship](https://github.com/agent-sh/ship) - PR workflow from commit to production\n\n## License\n\nMIT\n",
  "bytes": 4087,
  "sha": "d97cca49acb25fcc67f3134e0be9eab6aea3ba44d3496626a107892f74c48738",
  "repo_slug": "agent-sh/audit-project",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_agent_sh_audit_project_audit_project_497932c7/readme"
}