By garygentry
Authoring, review, guidance, and loop-operation skills for rauf autonomous coding loops. rauf owns the backlog contract; these skills serve both the repo-wide ad-hoc flow and feature-pipeline tools that delegate to them.
Author a high-quality rauf backlog.json — well-scoped, verifiable items for the rauf autonomous coding loop. Use this skill when the user asks to "create a rauf backlog", "author a backlog", "build a rauf backlog", "populate backlog.json", "generate the .rauf/backlog.json", "add backlog items", or "create the rauf task list". Serves BOTH the repo-wide ad-hoc flow AND feature-pipeline tools (e.g. feature-forge) that delegate backlog authoring here — parameterized by a target backlog directory (default `<project>/.rauf/`, or `--backlog <dir>` for feature/multi-backlog setups). Do NOT trigger for general planning or work breakdown that doesn't specifically mention a rauf backlog.
Operate the rauf CLI to drive an autonomous coding loop — the run → observe → recover lifecycle. Use this skill when the user asks to "run the rauf loop", "start the loop and watch it", "kick off rauf on this project", "supervise a rauf run", "check the loop status", "why is the loop paused/blocked/stuck", "recover an interrupted rauf loop", "resume the loop", "the loop died — fix it", or "what does rauf exit code N / status X mean". Also use when an agent needs to drive rauf programmatically (machine surfaces: `--ndjson`, `status --json`, `events.ndjson`). Do NOT trigger for authoring or QA-ing backlog items (use author-backlog / review-backlog), for fixing the installed `.rauf/RAUF.md` guidance (use review-rauf-guidance), or for behaving AS a loop iteration (that per-iteration contract lives in the project's `.rauf/RAUF.md`).
Review and QA an existing rauf backlog.json — a second-opinion audit of coverage, scoping, dependency sanity, acceptance-criteria quality, and enum correctness. Use this skill when the user asks to "review a rauf backlog", "review the backlog", "QA the backlog", "audit backlog.json", "check backlog against spec", "validate backlog items", "review backlog quality", or wants "a second opinion on the backlog". Serves both the repo-wide backlog (`<project>/.rauf/backlog.json`) and feature/multi-backlog setups (`--backlog <dir>`). Do NOT trigger for creating new backlogs — use author-backlog for that.
Review and update the rauf-installed guidance file (.rauf/RAUF.md) in a rauf-managed project to ensure it's accurate for the target project. Use this skill when the user asks to "review rauf guidance", "check rauf config", "audit rauf files", "update RAUF.md", "fix my rauf setup", or asks whether .rauf/RAUF.md is correct for their project. Also use when the user reports that the rauf loop is using wrong verification commands, missing project context, or behaving as if it doesn't understand the project.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Autonomous coding loops, managed.
Rauf installs, manages, and monitors AI coding loops across your local projects. Define a backlog, start the loop, and let Claude Code ship work items autonomously — with full visibility through a CLI and web dashboard.
Self-hosted from day one. Rauf built itself — every backlog item in this repo was implemented, verified, and committed by its own loop. The screenshots below are rauf managing rauf.
Each iteration produces one of three exit signals:
| Signal | Meaning | What happens |
|---|---|---|
RAUF_DONE | All acceptance criteria passed | Item marked done, loop continues |
RAUF_BLOCKED | Missing dependency or unclear spec | Item paused, loop retries or skips |
RAUF_NEEDS_HUMAN | Requires a decision or API key | Loop pauses for human input |
state.json with log-parsing fallbackbun build --compile (CLI + server + frontend + templates)Rauf is the default loop runner for feature-forge, an agent-agnostic spec-and-backlog pipeline that runs on Claude, Codex, Copilot, Cursor, and Gemini. feature-forge hands its generated backlog to a conforming runner; when no runner is configured it defaults to rauf. See feature-forge's README for the cross-agent install story and its per-agent setup docs.
Prerequisites: Claude Code CLI and Git (rauf spawns a Claude Code session each iteration and commits the result). Building from source also needs Bun 1.0+, pnpm 9+, and Node.js ≥ 22.
Building from source gives you the current version and is the supported path today:
git clone https://github.com/garygentry/rauf.git
cd rauf
pnpm install && pnpm build
bash scripts/install-global.sh # symlinks `rauf` into ~/.local/bin
rauf version # verify (~/.local/bin must be on your PATH)
Each tagged release also publishes self-contained binaries (no Bun/Node on the target machine), verified against the release's SHA256SUMS. This installs the latest published release, which may lag the source tree:
# Linux / macOS
curl -fsSL https://raw.githubusercontent.com/garygentry/rauf/main/scripts/install-binary.sh | bash
# Windows (PowerShell)
irm https://raw.githubusercontent.com/garygentry/rauf/main/scripts/install-binary.ps1 | iex
Set RAUF_VERSION=<tag> to pin a specific release instead of latest. macOS binaries are unsigned; if Gatekeeper blocks one downloaded via a browser, run xattr -d com.apple.quarantine ./rauf. See Releasing & Installing for details.
Once rauf is on your PATH (see Install above):
# Install rauf into an existing project
rauf install ~/workspace/my-project --yes
# Add a work item
rauf backlog add ~/workspace/my-project \
--title "Add user authentication" \
--type feature --priority 1 \
--ac "Login endpoint returns JWT" \
--ac "pnpm test passes"
# Start the loop
rauf loop run ~/workspace/my-project
npx claudepluginhub garygentry/rauf --plugin raufHarness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Reliable automation, in-depth debugging, and performance analysis in Chrome using Chrome DevTools and Puppeteer
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Design fluency for frontend development. 1 skill with 23 commands (/impeccable polish, /impeccable audit, /impeccable critique, etc.) and curated anti-pattern detection.