meta-skill-hygiene

Audit skill libraries for overlap, vague descriptions, bloat, and contradictions.

1|Updated Apr 3, 2026
One-click install
npx skills add https://github.com/TierOne-Studio/spa-velocity --skill meta-skill-hygiene-tierone-studio
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: meta-skill-hygiene
Source: https://github.com/TierOne-Studio/spa-velocity/tree/main/.ruler/skills/meta-skill-hygiene
Command: npx skills add https://github.com/TierOne-Studio/spa-velocity --skill meta-skill-hygiene-tierone-studio

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Skill libraries degrade over time: descriptions overlap and misfire, bodies bloat past readable length, rules contradict each other, and dead skills never trigger. This Skill provides a structured periodic audit that detects these issues and produces a prioritized cleanup proposal without making unapproved edits. ## Core Features & Use Cases - Six audit checks: Detects overlap between skill triggers, vague descriptions, body bloat (with a ~500-line split threshold and directory layout), contradictions with CLAUDE.md, dead skills, and missing disambiguation. - CLAUDE.md cross-coupling scan: Flags ADR references, file paths, code symbols, and subagent step citations that violate the layered-router principle. - Prioritized, approval-gated output: Produces a HIGH/MED/LOW issue list capped at ~5 high-priority items and stops before editing any file. - Use Case: After a month of adding skills, run the audit to find two skills whose descriptions both trigger on the same prompt, then merge them and sharpen the surviving description. ## Quick Start Ask the assistant to run a meta-skill hygiene audit over the skill library and produce a prioritized cleanup proposal for review.

Frequently Asked Questions about meta-skill-hygiene

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit an AI skill library for quality issues?

Run a structured audit that lists every skill, then applies checks for overlap, vague descriptions, bloat, contradictions, dead skills, and missing disambiguation. The output is a prioritized HIGH/MED/LOW cleanup list that a human reviews before any edits are made.

When should a SKILL.md file be split into multiple files?

Split when the body exceeds roughly 500 lines or when distinct sub-shapes like patterns, rules, and references emerge. The entry point stays under 200 lines and routes to REFERENCE.md, EXAMPLES.md, or pattern sub-files loaded on demand.

How do I fix two skills that trigger on the same prompt?

Overlapping descriptions cause inconsistent skill selection. Either merge the two skills into one, or sharpen each description with explicit 'Use when' and 'NOT for' clauses so the triggers become mutually exclusive.

Does the skill audit apply changes automatically?

No. The audit only proposes changes and stops before editing any skill file. The user reviews the prioritized list and approves each fix, since description edits alter triggering behavior and are treated as load-bearing.

What should CLAUDE.md contain relative to skills?

CLAUDE.md acts as a pure router pointing to skills and subagents by name only. It should not enumerate ADR numbers, file paths, code symbols, or subagent internal steps; those citations belong in the relevant skill or subagent file.