decision-council
Skill0 starsChanged 53 days agoMITView on GitHub
On-demand check of the user's own work or decision. Five isolated advisors judge it, then a fresh chairman agent merges them into one unbiased verdict with guidance. ONLY trigger when the user says: 'council this', 'run the council', 'check my thinking', 'review my decision', 'pressure-test this'. Do NOT auto-trigger on ordinary 'should I' questions, factual lookups, or creation tasks.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- tderoussel/agent-skillstheir own version, first seen Oct 20, 2026
- LeonardSEO/decision-counciltheir own version, first seen Jun 9, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- LeonardSEO/decision-council★ 1same name
- tderoussel/agent-skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
1. Brief (you, the main Claude): Write one neutral brief in two labelled parts.
- Very specific: line 18
Read only files the user named.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Decision Council
Goal: find out whether the user's decision or work makes sense, without one perspective biasing the result.
Ponytail ultra applies to the whole run; paste this Lean rule verbatim into every sub-agent prompt: Shortest answer that holds up. No preamble, no restating the question, no hedging, no filler. Stay under the word cap.
Steps
1. Brief (you, the main Claude). Write one neutral brief in two labelled parts. It is sent to every agent, so keep your own framing under 200 words:
- Decision: the decision, and the work itself when work is the subject: quote the load-bearing part verbatim (up to ~300 words, summarize beyond that).
- Context: key facts, then the user's reasoning as "User's stated reasoning:", with no lean and no opinion of yours.
Read only files the user named. Do NOT read memory/ or sweep the workspace. If the ask is too vague to brief, ask one question, then proceed. Then announce once, without asking permission: "Council: 6 agents."
2. Five advisors, one message, in parallel. Use the Agent tool with model: "sonnet". Each gets only: the brief, the Lean rule, and a single instruction built from its lens: "Review this by asking: [lens text]. Answer from the brief only. Do not use tools or read files. Max 100 words. Plain prose: no heading, label, or signature." The Outsider gets the Decision part only, so it has zero context.
Keep everyone blind. Each advisor prompt contains only the brief, the Lean rule and its lens question: no advisor names, and no mention of a council, chairman, or other reviewers.
| Advisor | Lens |
| Logic | Are the assumptions true? Does the conclusion follow from them? |
| Logistics | Time, resources, dependencies, order of steps. Can it actually be done? |
| Risk | What breaks? Failure modes? What is hard to undo? |
| Alternatives | A simpler or cheaper route? Opportunity cost? |
| Outsider | Zero context. What's unclear, or obvious to the user but not to anyone else? |
3. Chairman, one fresh agent (not you). You have seen the user's leanings in chat; a fresh agent has not. Give it ONLY the brief, the five answers in arbitrary order labelled Response A–E, and the Lean rule. Describe them only as "five independent reviews", with no mention of lenses or which came from which. If an advisor fails, pass the rest and say so. Never retry. Same model: "sonnet". It must not use tools, and must refer to points by content, never by letter. Ask for, in under 200 words:
- Verdict: Sound / Sound with fixes / Shaky, plus the one-line reason.
- Agree: points several responses reached independently.
- Clash: real disagreements, both sides, not smoothed over.
- Gaps: what all five missed, or flaws in the user's reasoning.
- Guidance: what to change, and the ONE first step.
It may overrule the majority if the reasoning supports it.
4. Present the chairman's output in chat as-is, with no opinion of yours added. No HTML, no files. Save a transcript only if the user asks.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.