Product Council

Metadata

• Trigger when: the decision is important enough that multiple explicit lenses would reduce blind spots or hidden risk.
• Do not use when: the task is a trivial copy tweak, obvious bug fix, or any decision where one clear answer is already enough.

Skill Purpose

Force meaningful disagreement across named lenses that optimize for different evidence and failure modes, then synthesize that disagreement into one recommendation.

Instructions

1. Start with one short decision brief covering problem, target user, current workaround, proposed move, key risks, and the actual decision to be made.
2. Choose 3-5 lenses and run them on the same evidence base. Use ~/.codex/skills/product-council/references/roles.md for role definitions and ~/.codex/skills/product-council/references/council-templates.md when a named template fits the decision better than the default council.
3. Synthesize the council into consensus points, disagreements, hidden assumptions, evidence that would change a mind, and one final recommendation: go, no-go, or proceed only after specific validation.

Non-Negotiable Acceptance Criteria

• Use named lenses, not celebrity cosplay or fake authority.
• All lenses reason from the same facts.
• The council produces one recommendation rather than a pile of unranked commentary.
• Evidence outranks performance; the exercise exists to reduce blind spots, not to sound grand.

Output

• The decision brief.
• Lens-by-lens findings plus explicit disagreements.
• A final recommendation with the key gate or proof needed to move forward.
• Next skill options (only if needed): $idea-validation — get real buyer/pain/proof signal before committing; $product-shaping — structure the chosen product question into one framework; $spec-bundle — turn the chosen direction into implementation-ready artifacts; $adr-log — record the decision if it is now durable enough.