llms.txt git:20260830.43ecb3dA

Immutable. This exact content is served forever at /api/v1/blob/eb91c74262da5354.

# mergeCraft

> AI-powered PR review as a standalone, BYOK GitHub Action. Deterministic
> analyzers + LLM reviewer + read-only verifier; findings gate merges via the
> mergecraft-approval commit status. No SaaS; your repo, your keys.

## Setup

- README.md — landing page, numbered install, Example 1 workflow
- README.md#for-agents — For LLM / Agents (copy/paste setup prompts)
- docs/install.md — Python 3.11+ floor, Action vs CLI, Docker fallback
- AGENTS.md — cross-vendor agent guidance (consumer setup + this-repo development)
- skills/mergecraft/SKILL.md — Agent-Skills consumer skill (setup checklist + CLI)
- skills/README.md — per-harness install matrix and regeneration notes
- skills/harnesses.yaml — harness manifest (install paths + primary sources)
- skills/codex/mergecraft/SKILL.md — generated Codex Agent Skills package
- skills/cursor/mergecraft/SKILL.md — generated Cursor Agent Skills package
- skills/opencode/mergecraft/SKILL.md — generated OpenCode Agent Skills package
- skills/gemini-cli/mergecraft/SKILL.md — generated Gemini CLI Agent Skills package
- skills/openclaw/mergecraft/SKILL.md — generated OpenClaw Agent Skills package
- skills/hermes/mergecraft/SKILL.md — generated Hermes Agent Skills package
- skills/claude-code/mergecraft/SKILL.md — generated Claude Code Agent Skills package

## Docs

- docs/mcp.md — public MCP install (stdio six-tool profile, per-runtime JSON)
- docs/mcp-tools.md — generated public MCP tool reference (six-tool profile)
- docs/README.md — generated index from docs/manifest.yaml
- docs/cli.md — full mergecraft CLI reference
- docs/cli-examples.md — runnable offline CLI example trees (examples/cli/)
- docs/action-reference.md — action.yml inputs and outputs
- docs/authentication.md — providers, custom gateways, model chains, init→auth→review
- docs/agent-roster.md — agent roster, slots, multi-reviewer, trust snapshot
- docs/workflows.md — workflow examples, trust tiers, pull_request_target
- docs/trust-policy.md — operator trust knob (execution vs authority, selfReview levels)
- docs/glossary.md — plain-language definitions for landing-page jargon
- REVIEW-CHECKS.md — every check a review applies
- docs/ANALYZERS.md — shipped analyzer catalog
- docs/REVIEW-DOCTRINE.md — review philosophy and grading
- docs/TRACING.md — tracing sinks and Logfire/OTLP wiring

## Agent surfaces

- .github/copilot-instructions.md — GitHub Copilot pointer to AGENTS.md
- .claude-plugin/plugin.json — Claude plugin manifest (skill + slash commands)
- commands/mergecraft-setup.md — slash command: scaffold without auth
- commands/mergecraft-review.md — slash command: local mergecraft review