question-hardlisted

## When to load this skill The question involves novel architectural tradeoffs, multi-layer debugging, subtle correctness or concurrency reasoning, security or performance analysis with non-obvious interactions, or anything where a wrong answer is expensive and the reasoning chain is long. ## When not to load this skill If the question is routine, use `/question-medium` or `/question-easy` instead - reserve this tier for genuinely deep or high-stakes reasoning. ## Procedure 1. Restate the question internally as a precise problem, not as the user phrased it. If $ARGUMENTS is ambiguous on a point that changes the answer, ask one focused clarifying question before proceeding. Do not ask about points that do not change the answer. 2. Establish ground truth before reasoning. If the question concerns a specific file, project, or repo, read the relevant files. If it concerns the current state of a fast moving tool, framework, library, or API, verify against the lockfile or authoritative source. Do not answer architectural or version-sensitive questions from memory. 3. If the stack matters: the `<repo-context>` block carries it when this runs inline; forked (no hook injection reaches the subagent), so absent that block, identify the stack from the project's own config directly. Load the patterns skill matching it if the question is in that area. 4. Reason explicitly. Name the assumptions the answer rests on. Where there is a tradeoff, state both sides and the axis that decides b