skills-cleanup

Detect and remove duplicate or unused agent skills and publish the curated library to public GitHub repos.

4|Updated Jun 17, 2026
One-click install
npx skills add https://github.com/exiao/meta-skills --skill skills-cleanup-exiao
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: skills-cleanup
Source: https://github.com/exiao/meta-skills/tree/main/skills-cleanup
Command: npx skills add https://github.com/exiao/meta-skills --skill skills-cleanup-exiao

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Skill libraries accumulate duplicates, dead pinned skills, and stale cross-references over time, wasting context-window tokens and breaking cron jobs that reference deleted skills. This Skill provides the workflows to audit, consolidate, archive, and safely publish a runtime skill library to public GitHub repositories. ## Core Features & Use Cases - Budget and usage reporting: Run the bundled bulk-budget-report.sh script to list every skill with token cost, body size, usage count, and flags like UNUSED, HEAVY_DESC, and LARGE_BODY. - Duplicate consolidation: Detect near-duplicate skills, salvage unique content into the canonical skill, and archive the retiree through the curator CLI with recoverable snapshots. - Safe publishing: Port runtime skills to public repos with redaction greps, cross-skill reference validation, worktree discipline, and PR workflows that avoid stale-base reverts and force-push pitfalls. - Recovery playbooks: Restore deleted runtime skills from the public mirror, curator backup tarballs, or written specs, and repair broken references in SOUL.md and cron jobs. - Use Case: Your agent reports "Skill(s) not found and skipped" for a cron job. Use this Skill to determine whether the skill was consolidated or deleted, recover its content from the newest curator backup tarball, and repoint every stale reference. ## Quick Start Ask the agent to run a skill budget report and clean up duplicate or unused skills in the library, then publish the curated result to the public repo.

Frequently Asked Questions about skills-cleanup

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

FAQPage Schema
How do I find unused or duplicate agent skills in my library?▼

Run the bundled bulk-budget-report.sh script, which scans every SKILL.md for token cost, body size, usage count, and last-used date. It flags UNUSED preloaded skills, oversized descriptions, large bodies, and description pairs with 40% or higher Jaccard similarity.

How do I recover a deleted skill that cron jobs still reference?▼

First grep for the skill's distinctive content to rule out consolidation into a sibling skill. If truly deleted, restore from the public mirror repo, then curator backup tarballs under .curator_backups, and only use hermes-backup if its HEAD is recent.

What should I check before publishing skills to a public GitHub repo?▼

Strip runtime-only frontmatter like preloaded: true, run the redaction grep for personal names, credentials, and internal identifiers, and validate that every references/*.md pointer resolves. Cross-skill reference validation before the first push prevents multiple rounds of CI review churn.

Why does git push from a worktree silently update nothing?▼

A worktree created with git worktree add -b sets the branch upstream to origin/main, so plain git push targets main instead of your branch. Always push with the explicit refspec git push origin HEAD:refs/heads/<branch> and verify the remote branch advanced.

When should I consolidate two similar skills instead of keeping both?▼

Consolidate when two differently-named skills cover the same class, such as X versus X-cli. Diff both bodies, port any unique content into the canonical pinned skill first, then archive the retiree through hermes curator archive so it stays recoverable.