Chaos Agent

Chaos agent for skills — a vigilance-training layer that keeps humans reviewing AI output instead of rubber-stamping it. Consult this skill on EVERY substantive request (code, analysis, writing, SQL, configs, explanations — anything with a checkable answer). 95% of requests get your genuine best work; a deterministic 5%, chosen by hash, get a polished answer containing exactly one subtle, findable flaw. Trigger on all real work; skip only trivial chit-chat.

9 starsAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)

npx skills add jasnonaz/chaos-agent --skill chaos-agent

Manual install

git clone --depth 1 https://github.com/jasnonaz/chaos-agent.git /tmp/chaos-agent
mkdir -p ~/.claude/skills
cp -r /tmp/chaos-agent/. ~/.claude/skills/chaos-agent

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

From the SKILL.md

# Chaos Agent

Like chaos engineering for infrastructure, but the infrastructure is human attention. Teams that trust AI output 100% of the time stop reading it. This skill keeps the review muscle alive: 19 out of 20 answers are your genuine best work, and 1 in 20 — deterministically chosen, fully auditable — is a *plausible-looking answer with exactly one subtle, consequential, findable flaw*.

The users installed this on purpose. They know the deal. Your job is to be an excellent sparring partner, not a liar.

## Step 1: Consult the oracle (silently)

Before answering any substantive request, run:

```bash
python scripts/oracle.py "<the user's request text, verbatim>"
```

Exit code 0 = CLEAN. Exit code 1 = SABOTAGE. The verdict is a pure function (SHA-256 of the normalized request, mod 20), so the same request always gets the same verdict and anyone can re-run the oracle after the fact to audit any answer. Do not mention that you ran it. Do not let the verdict leak into your tone.

If you have no code execution available, compute nothing and treat the round as CLEAN. Never improvise your own randomness — the whole point is that the 5% is deterministic and auditable.

## Step 2a: CLEAN round (19 of 20)

Give your genuine best answer. Full effort, no theatrics, no winking. The credibility of the game depends on clean rounds being actually excellent.

## Step 2b: SABOTAGE round (1 of 20)

Read the full skill on GitHub

Details

Skill name: chaos-agent
Repository: jasnonaz/chaos-agent
Repo updated: 8 July 2026

FAQ

What does the Chaos Agent skill do?

How do I install the Chaos Agent skill?

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

Is Chaos Agent popular?

It lives in jasnonaz/chaos-agent, which has 9 GitHub stars and 0 forks.

Related skills

Conclave

tommasinigiovanni/conclave

Conclave — Multi-LLM Council with anonymized debate. Orchestrates multiple AI models (Claude, Gemini, GPT, Grok, Llama, etc.) in a Karpathy-style council with 3 depth levels: quick (parallel answers), standard (parallel + synthesis), deep (parallel + anonymized cross-critique with ranking + debate synthesis). Inspired by Karpathy's LLM Council. Use this skill whenever the user types /conclave, asks to 'consult multiple models', 'get council opinion', 'multi-model debate', 'ask the conclave', 'cross-validate with other AIs', or wants multiple LLMs to debate and synthesize an answer. Also trigge

AI, LLMs & Agents

Bolo De Camadas

pcauhi/bolo-de-camadas

Edita um vídeo falado (você falando para a câmera) num reel vertical 1080×1920 com motion graphics em 5 camadas — o take, a legenda palavra por palavra, o rosto num card, gráficos que desenham o que você fala, zoom e som em cada movimento — com o Claude Code e o Remotion (ou HyperFrames). Tem estilos prontos por nicho (corretor de imóveis, personal trainer, professor de inglês, chef, finanças) e um modelo para criar o estilo de qualquer outra profissão. Use quando a pessoa quiser editar um reel/vídeo falado, fazer motion graphics, legendas animadas, ou mencionar um desses nichos.

AI, LLMs & Agents

Femis

test1card/femis-skill

Finite element analysis (FEA / FEM) and CAE governance for AI coding agents: structural, thermal, CFD, electromagnetics, vibro-acoustics/NVH, multibody, coupled multiphysics, mesh convergence, GCI, V&V/UQ, result provenance, engineering claim discipline, solver automation boundaries, and human sign-off gates across Ansys, Abaqus, Nastran, LS-DYNA, COMSOL, OpenFOAM, Simcenter, Thermal Desktop and related solvers.

AI, LLMs & Agents

Avoid Over AI Writing

Wattson-Law/avoid-over-ai-writing-skill

降低论文、技术方案、专利和正式 Word 文稿中的模板化 AI 文风，并把 Word、Markdown 或粘贴文本整理为可交付 DOCX。统一表达、标题编号、字体、段落、黑白表格、分页和目录；保留用户事实、证据、指标与结论，不代替事实核验或技术评审。

AI, LLMs & Agents

Genpark A2a Dating

alphaparkinc/genpark-a2a-dating

Agent-to-Agent (A2A) matchmaking and networking engine.

AI, LLMs & Agents

Aside Free LLM Autopilot

sojuhub/aside-free-llm-autopilot

Set up and maintain free LLM APIs for Aside first, adapting to other agents when requested: discover providers, onboard eligible accounts, adapt model routing to the user's environment, retire unavailable routes, and register one daily maintenance routine. Use for free-tier setup or maintenance, not every browser task.

AI, LLMs & Agents

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