agent-audit
When reviewing, optimizing, or inheriting an LLM agent, systematically identify what's actually load-bearing vs. cruft. Instruments usage, ablates components, and strips the agent to minimum viable form. Answers: "What can we remove without degrading quality?" The goal is the leanest agent that still works.
npx skills add majiayu000/claude-skill-registry --skill agent-audit-elliotjlt-claude-skill-potions --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
# Agent Audit <purpose> Most agents are over-engineered. Built for edge cases that never happen. Stuffed with tools that never get called. System prompts bloated with instructions that don't affect behavior. Running on expensive models for tasks that don't need them. This skill treats agent optimization like dead code elimination: instrument, measure, identify what's actually used, strip the rest. The goal isn't "what could be useful" but "what's actually necessary." </purpose> ## When To Activate Trigger when: - Inheriting or reviewing an existing agent - Agent costs are higher than expected - Agent is slower than it should be - "Can we make this cheaper/faster without breaking it?" - Preparing for production deployment - Refactoring agent architecture ## The Audit Framework <framework> ``` ┌─────────────────────────────────────────────────────────────────┐ │ AGENT AUDIT │ ├─────────────────────────────────────────────────────────────────┤ │ 1. INSTRUMENT What's actually happening? │ │ 2. MEASURE How often? How much? How expensive? │ │ 3. ABLATE What breaks if we remove X? │ │ 4. STRIP Remove what's not load-bearing │ │ 5. VERIFY Does it still work? Run evals. │ └─────────────────
- When To Activate
- The Audit Framework
- Instructions
- Step 1: Instrument the Agent
- Step 2: Measure Baseline
- Step 3: Identify Audit Targets
- Step 4: Ablation Testing
- Step 5: Strip and Verify
- Output Format
- The Minimum Viable Agent Test
- NEVER
- ALWAYS
- Examples
- Example 1: Tool Audit
What does the agent-audit skill do?
When reviewing, optimizing, or inheriting an LLM agent, systematically identify what's actually load-bearing vs. cruft. Instruments usage, ablates components, and strips the agent to minimum viable form. Answers: "What can we remove without degrading quality?" The goal is the leanest agent that still works.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill agent-audit-elliotjlt-claude-skill-potions --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.