my-skills

CLAUDE.md

Repository of reusable Agent Skills (Claude Code / Codex / OpenCode). Follow these rules when working here.

Language

Repository layout

SKILL.md authoring rules

Follow the official Agent Skills best practices:

  1. Frontmatter: name and description only.
    • name: lowercase letters, numbers, hyphens; ≤ 64 chars; must match the folder name; no reserved words (claude, anthropic).
    • description: ≤ 1024 chars, third person, non-empty. State what it does and when to use it, with key terms first — this is what drives auto-triggering.
  2. Body ≤ 100 lines. This repo caps skills at 100 lines (stricter than the official 500) — be direct and dense. If it grows past that, cut to the essentials or move reference material into linked files (references/…, one level deep, never nested).
  3. Be concise and assumptive — write only what the model doesn’t already know; don’t explain basics.
  4. One default path, not five options — offer a recommended approach plus an escape hatch.
  5. Consistency & portability — one term per concept, a defined output format, forward-slash paths, and no time-sensitive info (version-conditional notes like “EF Core 10+” are fine; hardcoded dates are not).

These are review skills

Each skill reviews only the changed lines of a diff and outputs a findings table ordered by impact, with this severity legend:

🔴 Critical · 🟠 High · 🟡 Medium · 🔵 Low

Keep every skill consistent with that shape: Scope → grouped criteria → Output format table. List only what needs to change; ignore what is correct.

Adding a new skill

  1. Create <name>/SKILL.md following the authoring rules above.
  2. Add a row to the README ## Skills table.
  3. Add the skill’s name to the copy commands in all three install/*.md guides.