{
  "markdown": "# PR Preparation and Review Agent Skills\n\nA collection of [Agent Skills](https://agentskills.io) for preparing changes and\nautomating pull-request review workflows.\nCompatible with Claude Code, GitHub Copilot, Cursor, Windsurf, and 30+ other agents.\n\n## Skills\n\n### Prepare Review Commits\n\nTurns the current uncommitted diff into a logical sequence of validated,\nself-contained Conventional Commits, ready for pull-request review. Never\nrewrites history or pushes. Manual-only: it runs only when you invoke it\nexplicitly (`/prepare-review-commits`), never on the agent's own initiative.\n\n**Install:**\n```bash\nnpx skills add xpepper/pr-review-agent-skill/prepare-review-commits\n```\n\n[See skill README →](prepare-review-commits/README.md)\n\n---\n\n### Guided Flow Review\n\nReviews a PR, branch, commit range, or working-tree change collaboratively with\nits author. It follows the runtime flow instead of diff order, presents one\nevidence-led step at a time, suggests refactorings where the change introduces\nor worsens a code smell, and waits for the author to route every finding before\nchanging code. Decisions and progress are kept in a disposable,\nresumable review TODO.\n\n**Install:**\n```bash\nnpx skills add xpepper/pr-review-agent-skill/guided-flow-review\n```\n\n[See skill README →](guided-flow-review/README.md)\n\n---\n\n### PR Review Loop\n\nAddresses all open PR review comments one at a time using an opinionated, resumable workflow.\nWorks with any reviewer (human or bot).\n\nRuns in a single session by default. An experimental `--handoff` mode completes one unit\nof work per session and stops, so you start each fresh session yourself.\n\n**Install:**\n```bash\nnpx skills add xpepper/pr-review-agent-skill/pr-review-loop\n```\n\n[See skill README →](pr-review-loop/README.md)\n\n---\n\n### PR Review Grill\n\nWalks through open PR review comments collaboratively, one at a time, challenging\npremises and recording the user's decision before any code changes.\n\n**Install:**\n```bash\nnpx skills add xpepper/pr-review-agent-skill/pr-review-grill\n```\n\n[See skill README →](pr-review-grill/README.md)\n\n---\n\n### PR Description Review\n\nReviews and improves pull-request descriptions so human and agent reviewers\nreceive the end-to-end flow, external contracts, rollout constraints, accepted\nlimitations, and evidence needed for focused feedback. Read-only: it drafts an\nimproved description without publishing it.\n\n**Install:**\n```bash\nnpx skills add xpepper/pr-review-agent-skill/pr-description-review\n```\n\n[See skill README →](pr-description-review/README.md)\n\n---\n\n### Copilot Review Loop\n\nAn in-session Copilot-driven review loop: triggers Copilot review, addresses\nits feedback one comment at a time, and re-triggers Copilot until all critical\nissues are resolved or 2 cycles are complete. Runs within a single agent session.\n\n**Install:**\n```bash\nnpx skills add xpepper/pr-review-agent-skill/copilot-review-loop\n```\n\n[See skill README →](copilot-review-loop/README.md)\n\n---\n\n### Ralph Wiggum Loop\n\nThe [Ralph Wiggum pattern](https://ghuntley.com/ralph/): an external shell loop\nthat spawns a fresh agent session for each PR comment. Drop `CODE_REVIEW_PLAN.md`\ninto your project root and run the loop — each invocation fixes one comment, then\nstops. Works with any agent (Claude, Codex, Cursor, etc.).\n\n**Install:**\n```bash\nnpx skills add xpepper/pr-review-agent-skill/ralph-wiggum-loop\n```\n\n[See skill README →](ralph-wiggum-loop/README.md)\n\n---\n\n## Requirements\n\n- `gh` CLI (recommended) or a GitHub token for REST API fallback\n- Git\n\n## Development\n\n### Skill structure\n\nEach skill is a directory containing:\n\n```\n<skill-name>/\n  SKILL.md          # required — skill definition with YAML frontmatter\n  README.md         # recommended — user-facing documentation\n  references/       # optional — supporting reference documents\n```\n\nThe `SKILL.md` frontmatter must include at minimum `name` and `description`. See existing skills for the full set of supported fields (`license`, `compatibility`, `metadata`, `allowed-tools`).\n\n### Packaging\n\nAfter adding or modifying a skill, regenerate the `.skill` package:\n\n```bash\n./package-skill.sh\n```\n\nThe script auto-discovers all directories that contain a `SKILL.md` and packages each one. Generated `.skill` files are gitignored (build artifacts).\n\n### Testing a skill\n\nAfter pushing changes, verify a skill installs and removes correctly:\n\n```bash\n# Install (--yes skips interactive prompts)\nnpx skills add xpepper/pr-review-agent-skill/<skill-name> --yes\n\n# Verify the skill description appears and the install path looks correct, then remove\nnpx skills remove <skill-name>\n```\n\nNote: without `--global`, skills are installed project-locally into `.agents/skills/` (gitignored).\n\n## License\n\nMIT\n",
  "bytes": 4727,
  "sha": "3489d976c9d19ae1139f1d039a16c24673c6a9553d248ec0832d0b9e73945bf7",
  "repo_slug": "xpepper/pr-review-agent-skill",
  "fonte": "repo",
  "truncated": false,
  "api": "https://api.agentalog.com/api/listings/skl_xpepper_pr_review_agent_skill_pr_review__96f70e05/readme"
}