agent-tools

Build and maintain an LLM-owned markdown wiki inside a repository. Use when setting up a repo wiki, scaffolding a wiki root, refreshing wiki pages from repo sources, ingesting new sources, answering from the wiki, or linting for gaps, stale content, and missing cross-links.

Create or update a project's `Makefile` with standard developer targets. Use when the user asks for a Makefile, wants `make build`, `make test`, `make run`, `make check`, or `make help`, or wants plain `make` to print help by default. Inspect the repo first and wrap existing project commands instead of inventing a new workflow.

Pull GitHub PR review threads and conversation comments via the gh CLI, triage them into verified fixes, implement changes, push updates, and optionally resolve review threads or minimise outdated comments; also poll every 4 minutes (sleep 240) up to 3 times for new feedback.

Respond to and address PR review comments. Fetches PR context, tracks resolution progress, and supports a workflow for systematically addressing reviewer feedback with agent assistance.

Build reproducible feature demos as markdown documents using showboat and Rodney browser automation. Use when the user asks for a demo, walkthrough, proof the UI works, screenshots of a feature flow, or an executable markdown artifact showing what was built.

Run a local Claude Code PR ultrareview using the repository prompt and Claude CLI, then triage the returned findings and fix valid issues. Use when the user asks for a Claude ultrareview, local ultrareview, unpaid ultrareview-style review, Claude CLI review of a PR, or asks to replace paid/cloud ultrareview with a local workflow.

Use when the user asks to clean, prune, or tidy git worktrees, or when worktree count seems excessive, or stale audit-mirror temp worktrees are accumulating.