Guide · Agent skill
How to use the council agent skill: four-voice decisions
Updated · written with AI from the project's own documentation and checked against our catalogue data
The council skill is an agent skill for Claude that handles fuzzy decisions, such as monorepo vs polyrepo or ship now vs wait. Instead of giving you one answer, it gathers four viewpoints and summarises where they agree, where they clash and what path it recommends. This guide explains what the skill does, how to install it from the ZIP on our page, how to trigger it with plain requests, and when to choose a different tool.
Four-Voice Decision Councilaffaan-m/ECC · MIT Download the skill Quick start
- 1Download the ZIP. Download the skill ZIP from this page and unzip it. It contains SKILL.md.
- 2Place the folder. Move the unzipped folder into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the skill there instead.
- 4Describe a fuzzy decision. Name the options, the constraints and what success looks like, and ask for a council or a challenge to your leaning.
- 5Read the four voices. Check the Architect, Skeptic, Pragmatist and Critic positions before reading the verdict.
- 6Decide and record. Use the recommendation to make the call, and save the result only if it changes something real.
What the council skill makes the agent do
When you face a decision with several reasonable options, the skill makes the agent weigh in from four lenses: an architect (accuracy, maintainability and long-term impact), a skeptic (challenging assumptions and simplifying), a pragmatist (release speed, user impact and operational reality) and a critic (edge cases, downside risk and failure modes).
Three of the four voices run as fresh sub-agents that see only the question and relevant context, not the whole conversation. The skill says this is meant to prevent anchoring, so the agent's earlier reasoning does not bias them. The agent first writes down its own initial position, then reads the other voices, and finally gives a compact verdict.
- Consensus: where the voices align
- Strongest dissent: the most important disagreement
- Premise check: whether the skeptic challenged the question itself
- Recommendation: the synthesized path
Install the council skill from the ZIP
Download the ZIP file from this page and unzip it. The skill package contains a single file, SKILL.md. Place the unzipped folder in one of these locations: ~/.claude/skills/ to make it available across your projects, or a project's .claude/skills/ folder to limit it to that project.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project only: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
Trigger the council with natural requests
You do not need a special command. Describe a fuzzy choice and ask for other views, a challenge to your leaning, or a multi-perspective opinion. The skill is designed for cases where there is no obvious winner and you want the trade-offs made explicit before a go/no-go call.
Good triggers name the options and the stakes. Vague requests may cause the agent to ask one clarifying question first, which the skill recommends when the question is unclear.
- Should we go monorepo or polyrepo for our next three services?
- Convene a council on whether we ship now or wait for more polish.
- I'm leaning toward feature flags over a full rollout. Challenge that before I decide.
- Give me a four-voice view on cutting scope versus keeping the strategic breadth.
Example: a ship-now decision
The skill's own example asks whether to ship a product as alpha now or hold until the control-plane UI is more complete. A typical council might have the architect argue for structural consistency and avoiding a confusing surface, the skeptic question whether the UI is really the gating factor, the pragmatist ask what can ship now without losing trust, and the critic focus on support load, expectation debt and rollout confusion.
The point is not agreement. The verdict is useful because it exposes disagreement before you choose.
What the skill needs and how it runs
The source lists no requirements. The skill does not run scripts and does not use the network. It is a single instruction file, so there is nothing extra to install beyond the skill folder itself.
The skill covers decisions under ambiguity only. For verifying output, decomposing features into implementation steps or designing system architecture, it points to other skills. Its default is one round; for another round, ask a focused new question and include the earlier verdict only if it matters.
- Requirements: none listed
- Runs scripts: no
- Uses network: no
- Files: SKILL.md
Tips for better council verdicts
Give the question, the key constraints and what success looks like. The skill asks for this shape before it convenes the council. Keep sub-agent context compact: include only relevant snippets or constraints, never the full chat history.
Save the outcome only when the decision actually changes something. The skill advises against writing ad hoc notes to shadow paths such as ~/.claude/notes; for lasting lessons it points to knowledge-ops, /save-session for session memory, or direct updates to the relevant GitHub or Linear issue.
- State the options and the constraints in one prompt
- Share only code that changes the answer
- Do not hide dissent in the final summary
- Use a separate skill for checking whether output is correct
What people use it for
- Choosing between monorepo and polyrepo for a team's codebase
- Deciding whether to ship a release now or hold it for polish
- Choosing feature flags over a full rollout for a risky change
- Deciding whether to simplify scope or keep strategic breadth
- Testing your own leaning against a deliberate skeptical challenge before a go/no-go call
Frequently asked questions
How do I install the council skill?
Download the ZIP from this page and unzip it so SKILL.md sits in a council folder. Put that folder in ~/.claude/skills/ for all projects, or in a project's .claude/skills/ folder for one project. In the Claude apps, you can upload it under Settings → Capabilities → Skills.
Does the council skill need anything else installed?
The source lists no requirements. It does not run scripts and does not use the network, and its package contains only SKILL.md.
Is the council skill free?
The source gives the licence as MIT and does not list any price. Check the repository's licence file for the terms that apply to your use.
What should I use instead of the council for code review?
The skill says not to use it for code review or bug finding. It recommends code-reviewer for bugs and security issues, or santa-method for verifying that output is correct.
Why does the skill use sub-agents that don't see my conversation?
Three of the voices run as fresh sub-agents with only the question and relevant context, so the agent's earlier reasoning does not anchor them. This is the skill's anti-anchoring mechanism.
Why isn't my council triggering or giving a long answer?
Clear options and constraints work best. If the question is ambiguous, the skill asks one clarifying question first. Straightforward tasks and factual questions are not meant for the council, so the agent may answer directly instead.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use the council skill when you have a real trade-off and want your leaning challenged from several angles before a go/no-go call. Skip it for code review, straightforward implementation or questions that have a verified correct answer, since it adds cost without adding value there.
Download the skill