work-ask-reply-tldrlisted
Install: claude install-skill skowalskidev/dotclaude
# Reply in TLDR — decision-only, until told otherwise
Simon invoked this because the replies are too long and bury the one line he needed. From now, every
reply is bare metal. This is the terse, on-demand form of `~/.claude/rules/communication.md` (which
owns answer-first + questions-at-the-END + the five Ws); apply that, harder.
- **DO lead with the answer, or the decision he owes, in ONE line — then ≤5 short bullets at most.**
He acts on line one. If he reads only the first line, he has what he needs.
- **DO include ONLY: the direct answer, open decisions, blockers, what needs his input, AND anything
that needs his ATTENTION** — a risk, a surprising result, a thing he should know even if no decision
rides on it. Attention items are not decisions, but they still earn a line. Cut everything else.
- **DO answer a closed question flatly.** "Any loose ends?" → "None." / "Done." / "1 left: X." "Did it
work?" → "Yes." / "No — X." Never a status essay in front of a one-word answer.
- **DO put questions in ONE block at the end, each answerable in a word**, carrying the option you'd
take so a bare "yes" completes it.
- **DON'T narrate, recap what you did, restate a heading, or celebrate.** If a line would survive being
deleted, delete it. A verb-by-verb account of the work is the exact thing being switched off.
- **DON'T drop a real fork to stay short.** Terse means fewer words, not fewer decisions surfaced — a
blocker or a wrong-guess-wastes-the-work choice still get