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.

62 stars 2 forksAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add ElliotJLT/Claude-Skill-Potions --skill agent-audit
Manual install
git clone --depth 1 https://github.com/ElliotJLT/Claude-Skill-Potions.git /tmp/Claude-Skill-Potions
mkdir -p ~/.claude/skills
cp -r /tmp/Claude-Skill-Potions/skills/agent-audit ~/.claude/skills/agent-audit

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# 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.                │
└─────────────────────────────────────────────────────────────────┘
```
</framework>

## Instructions

### Step 1: Instrument the Agent
Read the full skill on GitHub

Details

Skill name
agent-audit
Path
skills/agent-audit/SKILL.md
License
MIT
Repo updated
12 February 2026

FAQ

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 the Agent Audit skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add ElliotJLT/Claude-Skill-Potions --skill agent-audit. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Agent Audit popular?

It lives in ElliotJLT/Claude-Skill-Potions, which has 62 GitHub stars and 2 forks.

Related skills

Battle Plan

62

ElliotJLT/Claude-Skill-Potions

Before ANY significant development task (new feature, refactor, integration, migration), run a complete planning ritual by orchestrating other skills in sequence: rubber-duck (clarify scope) -> pre-mortem (assess risks) -> eta (estimate time) -> final confirmation. Do not start coding until the battle plan is approved.

Software Development

Breadcrumbs

62

ElliotJLT/Claude-Skill-Potions

Leaves notes for future Claude sessions in .claude/breadcrumbs.md. At session end or after significant work, records what was tried, what worked, what failed, and context for next time. On session start, reads breadcrumbs to understand history. Session-to-session memory without manual updates.

Productivity & Notes

Careful Delete

62

ElliotJLT/Claude-Skill-Potions

Before any destructive or irreversible operation (rm -rf, DROP TABLE, DELETE, force push, branch deletion, truncate), orchestrate a safety cycle: assess blast radius → confirm explicitly → document what happened. Chains pre-mortem and you-sure with built-in documentation. No dangerous operation executes without ceremony.

Git & Code Review

AI Draw

24

stone-yu/ai-draw-skill

单一 Skill 整合 PPT 演讲稿生成(36 主题 / 31 布局 / 27 动画 / 20 特效 / 15 全 deck 模板)与画图能力(7 种类型 × 8 主题)。两个平等的顶级模式:`--mode ppt` 输出可分享的演讲 HTML;`--mode single` / `--mode site` 输出技术图。`/ai-draw <需求>` 自动识别意图,模糊时询问。Trigger keywords (中+EN):画 / 画图 / 画一个 / 给我画 / 帮我画 / draw / diagram / chart / 架构图 / 流程图 / 时序图 / 知识图谱 / 思维导图 / 类图 / ER 图 / 演讲 / 分享 / 分享稿 / 讲稿 / 逐字稿 / PPT / deck / slides / keynote / 幻灯片 / pitch / 周报 / 课程 / 课件 / 小红书图文 / talk / presentation.

AI, LLMs & Agents

Redneck Engineer

44

Jakeschincariol/unhinged-claude-skills

Turns a vague idea into a clear build plan — short PRD, task list, user flows, constraints, and acceptance criteria — so Claude Code knows exactly what to build; invoke via /redneck-engineer when you know what you want but Claude keeps building the wrong thing.

AI, LLMs & Agents13

Brain MCP Setup

43

inite-ai/inite-brain-service

Walk a developer through connecting a fresh MCP client (Claude Desktop, Cursor, Goose v2, Aider, Continue.dev, n8n, or a raw @modelcontextprotocol/sdk client) to the INITE Brain service. Covers obtaining an API key, the per-tenant URL shape, config snippets per client, the scope matrix for the full tool surface (up to 32 built-in tools + 2 resources), and the smoke test. Use when the user says "add brain MCP", "connect brain to Claude", "set up brain for Cursor", or names any MCP-capable client.

AI, LLMs & Agents

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.