Matt Pocock
3314257466
aihero.dev renders an install widget above every skill page. Each page then repeated the same commands in its body, and the two copies had drifted: the widget uses the current `npx skills@latest ...` wording while the hand-written blocks mostly carried the older bare `npx skills ...`. Pages showed the right command and a stale one together. Delete the block from all 25 pages and record the rule in .agents/writing-docs.md: install wording belongs to the site, not the page. The template no longer carries a Quickstart, and .agents/install-block.md notes that docs pages are not one of its consumers.
3.5 KiB
What it does
wait-what is the fire extinguisher for a message that didn't land. You fire it the moment you lose the thread, and the agent re-pitches what it just said: a little of the context you were missing, plain English, and the vocabulary from your project's CONTEXT.md.
It is three lines long, and that is the design rather than an unfinished draft. Skills that fight verbosity fail by growing — a four-hundred-line concision skill still leaves the model verbose, because the model reads the volume, not the plea. This one carries a single precise leading word and nothing else.
When to reach for it
You invoke this by typing /wait-what — the agent won't reach for it on its own, and it shouldn't: only you know when you stopped following.
Reach for it the second you notice you're skimming — the agent has drifted into jargon it invented, stacked five acronyms, or explained a decision whose premise you never saw. It's a one-shot corrective on the conversation you're already in. To stop the jargon arriving in the first place, use grill-with-docs instead, which builds the shared language upfront.
The name is the mechanism
The leading word is wait. Not "be concise" — an instruction about the agent's output, which the model satisfies by clipping words and losing you further. Wait is about your state: it says comprehension failed here. An agent that hears "be brief" writes telegrams. An agent that hears "wait, you lost me" backs up and explains.
That distinction is the whole skill. Every popular fix for verbosity — /tldr, /no-fluff, /talk-normal — names the output, so the model over-corrects into a caveman register that's shorter and no clearer. Naming the listener asks for both halves at once: fewer words and the context you were missing.
The re-pitch is deliberately vague about its own scope. It says re-pitch that, not "that last message", because what lost you is usually bigger than one paragraph — the agent decides how far back to go.
It plugs into the language you already have
The body reuses the leading words already sitting in your global CLAUDE.md and your project's CONTEXT.md: ASD-STE100 Simplified Technical English for the register, ubiquitous language for the nouns. Skill, CLAUDE.md and CONTEXT.md reach for the same tokens, so firing it isn't a new instruction — it's a reminder of one the agent already agreed to.
If you have no CONTEXT.md, it still works; you just lose the domain-vocabulary half.
It's working if
- The re-pitch is shorter and clearer, not shorter and blunter.
- It adds the premise you were missing rather than just deleting words.
- Project nouns replace invented ones — the terms in your
CONTEXT.mdcome back. - You can fire it twice in a row without it degrading into terseness.
Where it fits
wait-what is a reach-for-it-anytime standalone — it sits inside whatever conversation you're already having, in any skill, at any point. It's the extinguisher; grill-with-docs is the sprinkler system, because a shared language agreed upfront is the real cure for jargon, and domain-modeling is what you reach for when the words themselves are the problem rather than one bad message. When you're unsure which skill fits the moment, ask-matt routes you.