Compose canonical grilling with domain-modeling.
If slash commands are available, /grilling with /domain-modeling is equivalent.
Otherwise load and follow both linked skills directly:
- Use
grillingfor the interview loop: rounds of frontier questions, a recommended answer for each, wait for feedback, and do not enact the plan until the user confirms shared understanding. - Use
domain-modelingfor durable language: inspect existing context docs and code, challenge fuzzy terms, updateCONTEXT.mdonly when terms are resolved, and create ADRs only when the ADR criteria are met.
Interview ownership remains with grilling; documentation is subordinate. Do not create
docs before the confirmation gate or just because this variant is active. Create or update
them lazily only when the confirmed conversation produces durable domain language or a
genuinely important decision.