RI Brew Guy’s AI Agent Skills

Skills I’ve built for AI coding tools. They’re just markdown with a bit of YAML frontmatter, the SKILL.md format that Claude Code, Gemini CLI, OpenAI Codex, GitHub Copilot CLI, and Cline all read natively. Other tools (Cursor, Aider) don’t have a native skills system, but you can point their rule files at these.

Each skill is shaped the same way. Write a draft, run it against its own test prompts, read the outputs side by side with a baseline, revise. The test prompts and per-case assertions ship with the skill in evals/evals.json, so “the skill works” means something you can actually measure instead of vibes.

Quick start 

/plugin marketplace add ribrewguy/agent-skills
/plugin install rest-api-design@ribrewguy-skills

That’s the Claude Code path. For other tools (Gemini, Codex, Copilot, Cline, Cursor, Aider), see Install.

What’s inside

  • rest-api-design: design and review HTTP REST APIs. Resource-oriented URLs, PATCH for state transitions, domain-expressive error codes, flat error envelopes, idempotency, content-type negotiation, typed contracts across TS/Python/Go/Rust.
  • structured-code-review: a review-only output format for code reviews. Source-of-truth-aware preamble, severity-tagged findings, file:line citations, no-findings-still-formal. Composes with domain-review skills.
  • task-handoff-summaries: three structured report formats (implementation summary before commit, worker handoff to orchestrator, closeout after completion). Hard rules against using the summary to mask incomplete work.
  • cross-agent-review: workflow for routine cross-vendor agent peer review (Claude reviews Codex’s work; Codex reviews Claude’s). The handoff package with self-assessment redacted, the cold-review discipline, the disagreement protocol.
  • multi-agent-git-workflow: git discipline for multi-agent work. Worktree-per-agent topology, orchestrator/worker roles, merge authority, acceptance/rejection rules, plus universal commit discipline (Conventional Commits, mandatory task ID, co-author line, UAT gate, no silent amends).
  • branch-promotion-discipline: the layer above multi-agent-git-workflow. 3-tier develop to uat to main promotion, per-tier CI gate matrix, source-ref enforcement, hotfix flow with forward-merge, branch protection ruleset, pre-commit hook setup.

The full catalog lives at Skills.

Why a skill collection?

A skill captures the opinionated part of how I work. The conventions and rules that don’t show up in any framework’s defaults but that I keep re-explaining to AI assistants across projects. Distilling them into a SKILL.md once and pointing every tool at the same file means I (and my future self, and anyone else using these) don’t re-litigate the same patterns every conversation.

The skills here are extracted from real policies/development/ directories I maintain in production projects. Each one passed an evaluation pass before it shipped: a 4-case test set with per-assertion grading, comparing skill-loaded output against a baseline.

Where to go next

  • Install: set up the skills for your AI tool of choice.
  • Skills: browse the catalog, see what each one covers, link out to the canonical SKILL.md.
  • Evaluations: every skill’s eval results, with side-by-side outputs and per-assertion grading. The receipts for “the skill works.”
  • Concepts: the cross-cutting ideas these skills share. The canonical-home symlink pattern, dependency tiers, severity ladders, composition over absorption.

This site is also the docs for contributors who want to add their own skills to the collection. See Skills → Contributing a new skill.