Stats
Actions
Tags
From dont-do-that
Operator explicitly marks the turn as a question, not a request, even when phrased as "fix X". Read-only tools only: file reads, search, and read-only shell commands. No file-edit tools or mutating shell commands. Name obvious fixes without applying them.
How this skill is triggered — by the user, by Claude, or both
Slash command
/dont-do-that:just-a-questionThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This is an explicit lock, not a blanket rule for every sentence ending in a
This is an explicit lock, not a blanket rule for every sentence ending in a
question mark. Use it when the operator says /just-a-question, "just a
question", "read-only", "do not change anything", or equivalent.
Do not apply this skill to QA/status checks on work you are currently driving: "is CI green?", "do the screenshots still match?", "does the PR body still describe the branch?", "is the release note still correct?", or similar. Those checks are part of the deliverable. Investigate with read-only tools first; if the answer is "no", the stale or broken artefact is a blocker, not merely an obvious future fix. Switch back to the normal workflow and fix it within the usual gates.
Stale PR evidence is never an acceptable final state for an active PR you are preparing for review. Regenerate, update, or remove stale screenshots, videos, PR body claims, and linked evidence before handing the PR back as ready.
A back-and-forth may follow. When the operator's framing shifts from question to request, say so in one line before the first mutation ("Reading this as a request for change now").
npx claudepluginhub epologee/laicluse-agent-fieldkit --plugin dont-do-thatCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.