← ClaudeAtlas

skills-reviewerlisted

Reviews Agent Skills / SKILL.md packages for quality, spec compliance, trigger clarity, examples, references, eval coverage, bloat, conflicts, and risky instructions. Use when asked to audit, review, improve, score, or validate an agent skill.
itzikiusa/otto_os · ★ 0 · AI & Automation · score 76
Install: claude install-skill itzikiusa/otto_os
# Skills Reviewer Use this skill to review an Agent Skill package before sharing, installing, publishing, or committing it. The goal is not only to validate syntax. The review must answer: **will this skill be reliably selected, safely executed, easy to maintain, and useful under real tasks?** ## Inputs this skill can review - A skill folder containing `SKILL.md`. - A single `SKILL.md` file. - A proposed skill design, draft, or diff. - A skill package that also includes `scripts/`, `references/`, `assets/`, `agents/`, examples, or eval files. ## Review workflow 1. **Identify the review target.** Confirm the skill root, whether `SKILL.md` exists, and which bundled files matter. 2. **Check Agent Skills structure.** Validate frontmatter, required fields, naming, directory layout, relative file references, and package portability. 3. **Evaluate activation quality.** Review `name` and `description` for clear trigger terms, scope boundaries, non-overlap with other likely skills, and no marketing fluff. 4. **Evaluate instruction quality.** Look for a focused workflow, imperative steps, explicit inputs and outputs, deterministic decision rules, edge cases, and safe fallback behavior. 5. **Evaluate examples and references.** Require examples that demonstrate expected behavior, counterexamples that show when not to use the skill, and references that are focused, cited, and loaded on demand. 6. **Evaluate scripts and assets.** Scripts should be deterministic, small, documented, de