agentleFS
hsigstadhsigstad/research-kit/skills/style-revise/SKILL.md

style-revise

Skill0 starsChanged 4 months agoMITView on GitHub

Revise a rough draft paragraph by fixing idea flow before wording, then offer alternative shapes. Use when the user shares unpolished prose and asks for comments, says a paragraph "feels awkward", wants "better flow", restructuring, wording alternatives, or help finding the right shape for ideas they've already drafted. For auditing a near-final document against the style guide use /style-review; for a fast mechanical AI-tell scan use /style-check.

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 12 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈12

    Distinctive · estimate

    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
  1. Very specific:

    Use it when the user hands you a messy paragraph and wants to find the right shape — not when they want a compliance pass on a document …

    line 13
  2. Distinctive:

    Don't add or drop ideas by default.

    line 24
  3. Distinctive:

    Treat an awkward sentence as evidence of an idea-flow problem, not a bad-English problem.

    line 28
  4. Distinctive:

    If working in the user's workspace, also load the overlay (banned phrases, referee carve-out, author tics):

    line 44
  5. Distinctive:

    If you can tell the paragraph's section type (intro, results, theory, abstract, …), load the per-type guide too:

    line 51
  6. Distinctive:

    Read the target prose and enough context to understand the local argument.

    line 63
  7. Distinctive:

    If the idea structure is what's blocking good flow, name that first.

    line 76
  8. Distinctive:

    Don't try to fix structural confusion with surface polish.

    line 77
  9. Distinctive:

    Put the main point where it can organize the paragraph.

    line 84
  10. Distinctive:

    Add a transition only after the underlying relation is clear.

    line 88
  11. Distinctive:

    Before presenting a revision, run it through the mechanical linter so you don't hand back prose the rest of the stack would immediately …

    line 94
  12. Distinctive:

    If a proposed rewrite trips a rule, fix it before showing it.

    line 101
  13. Very specific:

    If the user wants the edit applied, edit the source file, not a generated build output.

    line 107
  14. Distinctive:

    Keep line-level proofreading comments out unless they materially affect flow.

    line 121

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

style-revise

Take a rough draft and get the ideas into the right order before touching the words. The other style skills audit finished prose against the rulebook; this one reshapes an unfinished draft. Different job, earlier in the pipeline.

Use it when the user hands you a messy paragraph and wants to find the right shape — not when they want a compliance pass on a document that's already close to done. That's /style-review.

Stance

Assume the user has the right ideas and needs the best structure for them. Revise aggressively at the level of order, emphasis, sentence boundaries, transitions, and paragraph shape — but preserve the substantive claims, notation, and technical language.

Don't add or drop ideas by default. Add a missing idea, cut one, or move one out of the paragraph only when the paragraph can't work otherwise — and say so explicitly when you do.

Treat an awkward sentence as evidence of an idea-flow problem, not a bad-English problem. It's usually awkward because it's doing two jobs at once, because the sentence before it didn't set it up, or because the paragraph's logical order isn't right yet.

Workflow

1. Load the rulebook first

Your rewrites have to survive the same standards the rest of the stack enforces. Before proposing anything, load the general baseline:

research-kit/rules/writing_style.md

If working in the user's workspace, also load the overlay (banned phrases, referee carve-out, author tics):

research/rules/writing_style.md

If you can tell the paragraph's section type (intro, results, theory, abstract, …), load the per-type guide too:

research-kit/rules/writing_style/<type>.md

This is what keeps a "smoother" rewrite from quietly reintroducing an AI-tell or a banned phrase.

2. Diagnose flow before wording

Read the target prose and enough context to understand the local argument. Identify the paragraph's job: the claim, contrast, mechanism, qualification, implication, or transition it needs to deliver.

Then look for the structural problems, in this order:

  • Ideas in the wrong order.
  • The main claim buried below the setup.
  • A sentence doing multiple logical jobs.
  • Missing connective tissue between ideas.
  • A qualification that lands before the claim it qualifies.
  • A final sentence that belongs earlier.

If the idea structure is what's blocking good flow, name that first. Don't try to fix structural confusion with surface polish.

3. Rearrange, then smooth

Repack the existing ideas into a clearer sequence. Only once the order is right, make the language cleaner and more economical:

  • Put the main point where it can organize the paragraph.
  • Split sentences doing multiple logical jobs; combine ones whose separation makes the logic choppy.
  • Move qualifications after the claim, unless accuracy demands otherwise.
  • Add a transition only after the underlying relation is clear.
  • Replace vague connectives with the actual relation: contrast, mechanism, implication, caveat, example, consequence.

4. Lint your own rewrite

Before presenting a revision, run it through the mechanical linter so you don't hand back prose the rest of the stack would immediately re-flag:

python3 ~/research/research-kit/tools/style_lint.py <tmpfile-or-source> --severity warning

If a proposed rewrite trips a rule, fix it before showing it. A rewrite that reads well but reintroduces a banned phrase is a regression, not an improvement.

5. Apply, if asked

If the user wants the edit applied, edit the source file, not a generated build output. Preserve notation, citations, [ns:key] tokens, \labels, author notes, and technical claims unless the user explicitly asks to change them. Don't commit automatically.

Output style

For a short interactive request, usually:

  1. A one- or two-line diagnosis of what's limiting the flow.
  2. Two or three revised versions, when alternatives actually help — e.g. conservative vs. more aggressive restructuring.
  3. A short note on the tradeoff between them.

Keep line-level proofreading comments out unless they materially affect flow. For a long document, give a structured revision report rather than inline rewrites, unless the user asks for direct edits.

When to use vs. the other style skills

SkillJobDraft maturityOutput
/style-checkmechanical AI-tell scananylinter findings
/style-reviewaudit against the style guidenear-finalrule-citing findings
/validate-sectionsection-scoped pre-submission gatenear-finalfindings + ledger
/style-revise (this)reshape idea flowrough / unpolisheddiagnosis + alternative rewrites

The dividing line: rough draft you're still shaping → /style-revise; document that's basically done and needs a compliance pass → /style-review. When the user says "polish," disambiguate by draft maturity, not the word itself.

Cost

One LLM pass over a paragraph or short passage, plus a linter run on the rewrite. Cheap for interactive paragraph-level work. For a whole document, prefer /style-review (which is built for that scope) unless the user specifically wants structural reshaping rather than a guide-compliance audit.

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.