skills/claude-code-audit-workspace/iteration-1/eval-2-casual-phrasing/with_skill/outputs/audit/drafts/skills/skill-dead-code-sweep/SKILL.md
Reviews ~/.claude/skills/ against the last N days of actual session transcripts to identify skills that never triggered and are paying context-noise cost for no value. Produces a disable/archive recommendation per skill. Use when the user has 30+ skills installed and wants to reduce trigger-time noise, or when the user asks to "clean up my skills", "find unused skills", or "which skills never fire".
npx skillsauth add petekp/claude-skills skill-dead-code-sweepInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
The audited user has 40+ skills installed under ~/.claude/skills/. In a 25-session sample covering 30 days, the Skill tool was invoked 41 times total, dominated by:
circuit:handoff — 91 invocationscircuit:run — 34 invocationscodex — 11 invocationscircuit:explore — 4 invocationsThe rest — api-design-patterns, clean-architecture, ubiquitous-language, typography, improve-codebase-architecture, literate-guide, formal-verify, grill-me, dogfood, swift-apps, swiftui-expert-skill, deep-research, deepwiki, seam-ripper, ai-sdk, agentation, browser-use, agent-browser (duplicate of browser-use), process-hunter, and others — did not fire in the sample.
Each installed skill contributes to the skill-list system reminder, which is re-sent every turn. With 40+ skills that list is large, costs tokens, and dilutes trigger accuracy: a skill with a vague description can swallow triggers that belong to a better-matched one.
~/.claude/skills/*/SKILL.md.python ~/.claude/skills/claude-code-audit/scripts/inventory.py to list sessions from the last 30 days.Skill tool invocations (the skill field of the tool_use) and skill_invocations from the extracted JSON if present.invocations_in_window, projects_invoked_in, days_since_last_invocation.A report with four sections:
~/.claude/skills-archived/.security-review, schedule). Confirm with the user first.development
Draft short, plainspoken notes in the author's voice that help reviewers understand non-obvious choices, boundaries, and preserved behavior in the author's own pull request or local diff. Use when the user asks to self-review, annotate, or add reviewer context to their PR or changes. Draft locally when no PR exists, and post approved notes as one GitHub review when a PR does exist. Do not use for reviewing someone else's PR, writing code comments, explaining code generally, or drafting a PR description. Never post without explicit approval.
tools
Design and build pure-CSS (zero-JavaScript) Tailwind CSS v4 plugins of unusual depth and craft. Use when the user wants to create, architect, or refine a Tailwind utility plugin or CSS effect — e.g. "make a tailwind plugin", "build a tw-* plugin", "a CSS-only shimmer/fade/glow/grain/noise utility", "tailwind v4 @utility", "package this effect as a plugin", or wants an effect with surprising visual depth (gradients, masks, filters, SVG filter tricks, scroll-driven animation). Pairs deep CSS/SVG technique research with a bespoke tuning workbench for dialing the effect in. Inspired by tw-fade and tw-shimmer.
content-media
Create clear, polished before-and-after screenshots for a GitHub pull request. Use when a UI change needs visual proof: capture matching states, crop to the relevant UI, stitch and caption one comparison image, attach it natively to the PR, and keep the image out of the repository.
testing
--- name: latent-potential description: First-principles, team-of-experts assessment of a software project that surfaces latent potential; underexploited assets, a sharper north star, missing high-leverage capabilities, better framing and messaging. Produces a prioritized, evidence-grounded report with cheap probes, a reframe candidate, a stop-doing list, and an honest skeptic's case. Use whenever the user wants fresh eyes on a project they have built: "what am I sitting on", "what could this be