| name | ask-risk |
| description | Planning risk interview for finding assumptions, contradictions, missing requirements, privacy issues, UX gaps, education-specific gaps, and implementation ambiguity before building. Use when a plan feels underspecified or when rework would be costly. |
Ask Risk
Overview
Use this skill to find what the AI might misunderstand before implementation starts.
Focus on questions that prevent rework, not theoretical completeness.
Risk Categories
Review the plan for:
- Goal mismatch
- User/audience ambiguity
- Scope creep
- Missing non-goals
- Unclear success criteria
- Unspecified platform or device
- Visual direction ambiguity
- Data, API, or integration uncertainty
- Privacy or safety risk
- Accessibility/readability risk
- Education-specific assessment or feedback gaps
- QA path missing
Interview
Ask only high-impact questions. Prefer 3-7 targeted questions.
Use this pattern:
- State the risk in plain language.
- Explain what could go wrong if unanswered.
- Ask one concrete question or offer A/B/C choices.
Output
Return:
- Risk ledger
- Blocking questions
- Non-blocking assumptions
- Recommended changes to the implementation prompt
- Change log: one line per changed decision, in the form
Before -> After (addresses: <risk>). Include only what actually changed, not the full plan.
Move blocking questions into the running plan's Open Questions, non-blocking items into Assumptions, and apply the recommended changes directly to Confirmed Decisions before handing off to ask-prompt. Pass the change log along with the updated plan — it is how ask-prompt checks that nothing found during risk review got lost on the way to the final prompt.
