matiinchauspe yun-grill-plan
Interview the user relentlessly about every aspect of this until you reach a shared understanding. Walk down each branch of the decision tree, resolving dependencies between decisions one by one. For each question, provide your recommended answer.
Ask the questions one at a time, waiting for feedback on each before continuing. Asking multiple questions at once is bewildering.
Facts vs decisions
If a fact can be found by exploring the environment, look it up rather than asking — the filesystem, tools, the codebase, and prior decisions from past sessions are all fair game. Before putting any decision to the user, recall prior work on the decision's keywords (follow skills/_shared/memory-convention.md): if it was already settled in a prior session, surface it ("you decided X before — still holds?") instead of asking fresh. Nothing relevant found → ask. Any recommended answer you attach is grounded in what the lookup turned up, not a lazy default.
Surface only decisions an implementer couldn't safely default on their own; skip trivia they'd just pick. That bound is what keeps the interview relentless without turning endless.
The decisions themselves are the user's. Put each one to them and wait for the answer.
As decisions resolve
Persist each settled decision as it lands — not at the end — following skills/_shared/memory-convention.md. The grilling front-loads the questions so the work afterwards runs autonomously; the record is what makes a later session (or a fresh agent) able to pick it up without re-asking.
Done
Do not act on the plan until the user confirms you have reached a shared understanding. The bar: an implementer could execute the result without asking a single question.
Attribution
Adapted for this workspace from Matt Pocock's grilling (github.com/mattpocock/skills, MIT). Change: facts-lookup extended to recall prior decisions, and settled decisions are persisted as they resolve — both via skills/_shared/memory-convention.md.