agentleFS
chuckplayerchuckplayer/claude-agent-pack/skills/what/SKILL.md

what

Skill8 starsChanged 35 days agoMITView on GitHub

Restate the immediately preceding assistant message in ASD-STE100 Simplified Technical English, keeping the domain's own vocabulary as approved Technical Names. Use when a reply was accurate but hard to read — dense, hedged, idiomatic, or pitched at a specialist. Every fact, qualifier, negation, count, identifier and refusal survives verbatim: a status word such as NOT RUN or unverified is a Technical Name and is never simplified away. Trigger this when someone says: what, say that in plain English, simplify that, restate that, what did you just say, put that in STE. Do NOT use to summarise or shorten — STE output is normally longer than its source. Do NOT use to translate domain terms into everyday words. Do NOT use on any message except the last one.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Write to these. They are the ones that change the text most.

    line 44
  2. Very specific:

    Keep the articles. the report, not report.

    line 51
  3. Distinctive:

    No participial phrase standing in for a clause.

    line 53
  4. Very specific:

    No idiom, no metaphor, no slang, no understatement.

    line 59
  5. Distinctive:

    Do not merge the source's points to hit a paragraph limit: Add a paragraph instead.

    line 109
  6. Distinctive:

    Do not fix the source: If the preceding message was wrong, say so in a new reply — restating it incorrectly-but-clearly is worse than …

    line 110
  7. Distinctive:

    The domain vocabulary is the reader's, not yours: If the conversation used a term, keep that term even where a more common word exists …

    line 116

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

What

Restate one message — the assistant's immediately preceding reply — in ASD-STE100 Simplified Technical English, using the domain's own terms as approved Technical Names.

This skill adds nothing and removes nothing. It changes register only. It is not a summary, not an explanation, and not a second answer. If the restatement would be improved by new information, that information belongs in a new reply, not in this one.

1. Resolve the target

The target is the assistant's last message in this conversation — the one directly above the invocation. Not a file, not a selection, not the whole thing so far.

ConditionAction
No preceding assistant messageStop and say so. Never restate the user's own message instead
The preceding message was itself a /what outputStop and say so. A second pass compounds loss and gains nothing
The preceding message is a bare question with no contentRestate the question. A question is a message
An argument is supplied (e.g. /what insurance)Treat it as the domain, per step 2

2. Fix the domain before writing anything

ASD-STE100 has two vocabularies, and this is what makes it fit for a specialist reader: a controlled general vocabulary of roughly a thousand approved words, each with one approved meaning — plus Technical Names and Technical Verbs, which are unlimited and drawn from the subject.

So the domain term is not the problem. It is the sanctioned part.

  • Take the domain from the argument when one is supplied.
  • Otherwise infer it from the conversation and name your inference in one line before the restatement, so a wrong guess is visible rather than silent.
  • Never translate a Technical Name into everyday words. subrogation does not become "getting the money back from the other party". binder does not become "temporary paper". The reader knows the term; it is the sentence around it that was hard.

3. Apply the rules

Write to these. They are the ones that change the text most.

Sentences

  • One instruction per sentence.
  • 20 words maximum in an instruction. 25 maximum in descriptive text.
  • Active voice. Use the passive only where the actor is genuinely unknown or irrelevant.
  • Keep the articles. the report, not report.
  • Simple tenses only — simple present, simple past, simple future.
  • No participial phrase standing in for a clause. Split it into two sentences instead.

Words

  • One term for one thing, every time. Never vary a term for style — a synonym reads as a second concept.
  • No idiom, no metaphor, no slang, no understatement. earned its keep becomes a statement of what it did. alarming becomes the fact that caused the alarm.
  • Noun clusters: three words maximum. Break longer ones with a preposition.

Paragraphs

  • One topic each. Six sentences maximum in a procedure, ten in description.
  • A sequence becomes numbered steps.
  • A warning or a caution goes before the step it applies to, never after.

4. The fidelity rule — this is the one that matters here

Simplification deletes qualifiers, and in this project the qualifiers are the content. A reply saying a check is NOT RUN, a claim is unverified, a line is informational and never a stop, or a gate refused carries its whole meaning in those words. A restatement that loses one has not been simplified. It has been falsified, and it now reads as more confident than its source.

So:

  • Every status word, negation, hedge, count, identifier, file path, command, work item id and error code is a Technical Name. Reproduce it exactly. Do not reword it, round it, or drop it.
  • Where a sentence cannot meet the word limit without losing a qualifier, split the sentence and keep the qualifier. The limit yields; the qualifier does not.
  • Where the source says something was not done, not measured, or not proven, the restatement says so in the same strength. could not be measured does not become was difficult to measure.
  • Where the source refuses something, the restatement refuses it. Do not soften a refusal into a preference.

A shorter output is a warning sign, not a success. STE trades length for clarity: short sentences, repeated terms, no compression. Expect the restatement to be longer than its source. If it came out shorter, something was dropped — find it.

5. Output

Emit, in order:

  1. One line naming the domain used and whether it was supplied or inferred.
  2. The restatement.
  3. A short Technical Names list — the domain terms and status words you carried through unchanged.

That third item is the check on this skill. It shows the reader which words were treated as vocabulary rather than simplified, so a term wrongly protected — or wrongly flattened — is visible without them holding both texts side by side.

Nothing else. No preamble, no offer of further help, no restatement of these rules.

Gotchas

  • This is a register change, not a reading level. STE is written for a qualified reader working in a second language, not for a beginner. Do not explain the subject.
  • Do not merge the source's points to hit a paragraph limit. Add a paragraph instead.
  • Do not fix the source. If the preceding message was wrong, say so in a new reply — restating it incorrectly-but-clearly is worse than leaving it.
  • A table stays a table. STE governs sentences. Reformatting a table into prose loses the structure that made it readable.
  • Code, commands, and identifiers are quoted, never simplified. They are Technical Names with no approved substitute.
  • The domain vocabulary is the reader's, not yours. If the conversation used a term, keep that term even where a more common word exists outside the domain.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.