clean-room · backnotprop — Claude Code skill for independent decisions
clean-room is a Claude Code skill that asks a fresh agent to assess a decision from a neutral brief, without the current conversation. Run it for an independent verdict.
no license 344 installs ★ 194 updated 2026-10-15
npx skills add backnotprop/bro --skill clean-room -a claude-codeWhat clean-room does
clean-room prepares a neutral brief and gives it to a separate agent to assess one decision. Keep each run focused on either whether to act or how to act. The brief can include confirmed facts, constraints, unresolved questions, and evidence; it should leave out opinions and hints about the preferred outcome.
The agent is asked to test assumptions, consider alternatives, and look for evidence against its conclusion. Its response must give a verdict, supporting reasons, the strongest opposing argument, and what could change the verdict. A rerun starts with a fresh agent and excludes earlier clean-room conclusions. Since the agent receives no conversation history, each run needs its own brief.
What it helps with
- get an independent view of a disputed decision
- separate whether-to-act from how-to-act questions
- test a conclusion against opposing evidence
- weigh alternatives, including taking no action
Example prompts for the agent
- Use clean-room to assess whether this feature belongs in the project.
- Get an independent verdict on switching libraries or keeping the current one.
- Compare the migration options and return a clean-room decision.
How to install clean-room
1Skills CLI (Vercel)
npx skills add backnotprop/bro --skill clean-room -a claude-code2Manually: put the files in place
~/.claude/skills/clean-room/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/clean-room/
1Skills CLI (Vercel)
npx skills add backnotprop/bro --skill clean-room -a codex2Manually: put the files in place
.agents/skills/clean-room/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add backnotprop/bro --skill clean-room -a cursor2Manually: put the files in place
.cursor/skills/clean-room/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install backnotprop/bro skills/clean-room2Manually: put the files in place
.github/skills/clean-room/SKILL.md1Skills CLI (Vercel)
npx skills add backnotprop/bro --skill clean-room -a gemini-cli2Platform CLI
gemini skills install https://github.com/backnotprop/bro --path skills/clean-roomManually: put the files in place
.devin/skills/clean-room/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add backnotprop/bro --skill clean-room -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about clean-room
How do I install clean-room in Claude Code?
Run npx skills add backnotprop/bro --skill clean-room -a claude-code. The skill file is at ~/.claude/skills/clean-room/SKILL.md.
What does clean-room do?
It gives a separate agent a neutral brief so the agent can assess a decision without the current conversation. The response includes a verdict, its reasons, and the strongest opposing case.
Why does clean-room use a fresh agent for each rerun?
Each run starts without the previous conversation or clean-room verdict. The agent receives a newly prepared brief for its assessment.
Similar skills
All skills for Agent setup: memory, context, orchestration →clean-room by other authors comes first.