/expert-council

Garden · not yet reviewed Beta, from the WDAI Build Cycle pilot Version 0.1.0

Vets an idea, plan, draft or decision from three to five relevant perspectives, each saying what it would do and what worries it, and ends with where they agree and disagree. Modes include a pre-mortem ("it failed — why?") and devil's advocate.

How to call it

Once it's installed, ask for it by name, for example "use expert-council on this" or type /expert-council . Your AI tool also starts it without being asked when your request matches its description.

When it starts: someone says "poke holes in this", "what am I missing", "stress-test this", "is this a good idea", "pre-mortem", or is about to commit to something consequential.

Connections

Practises

• Check the result
• Frame the outcome

Concepts it uses

• Sycophancy (garden)
• Human in the loop (garden)

Comes after

• /build-brief (garden)
• practice-coach (garden)
• /brainstorm (garden)

Try next

• /prove-it (garden)

Used in blueprints

• Ask AI to play devil's advocate on a team decision (garden)

From the field, after this

• draft-review-kit (garden)
• Wharton Generative AI Labs prompt library (garden)

The skill itself

This is the instruction text your AI tool reads, shown as written.

Expert council

Look at one idea from several perspectives, so the person sees the gaps before they cost anything.

How this skill behaves

They state their position first. The council doesn’t agree by default; each perspective looks for problems. “Just do it” runs the council straight away.

Steps

1. Their position first. Ask what they currently think and how confident they are, from 1 to 5. This stops a persuasive council from overriding their judgement without them noticing.
2. Choose a mode. Ask which they want, or suggest one:
   • Council (default): three to five perspectives that matter for this decision. Examples: the person it’s for, whoever pays, whoever has to run it, a domain expert, a skeptic. Suggest a set and let them swap any.
   • Pre-mortem: “It’s six months from now and this failed. Why?” They list reasons first; the council adds more.
   • Devil’s advocate: the strongest case against their position.
   • Attack the premise: if two fixes built on the same assumption have failed, test the assumption itself.
3. Each perspective says what it would do. Each one says what it would do with this idea, what it would need to see, and what worries it most. Keep each to a few lines, with no role-play voices.
4. Synthesise. Where do they agree? Where do they disagree, and why? What is the single biggest risk?
5. They decide. Ask what they’ll change, if anything, and whether their confidence moved. Record the decision and the reason in their words.

Output

A short table of perspectives, then where they agree, where they disagree and the biggest risk, then their decision and reason. Offer prove-it for any claim the council couldn’t settle.

---

Credit: Draws on Felix Rieseberg ('assemble a crew'), Gary Klein's pre-mortem, Advait Sarkar ('AI should challenge, not obey') and pstack principle-attack-the-premise (MIT, Lauren Tan).

Licence: Apache-2.0 (instructions) and CC BY 4.0 (teaching text). From AI Practice Commons by Women Defining AI.. Plain Markdown for agents: /skills/expert-council.md