agentleFS
guerrilla2799guerrilla2799/ops-and-scale-os/skills/writing-quality/SKILL.md

writing-quality

Skill1 starsChanged 3 months agoMITView on GitHub

Runs automatically on any prose this repo produces that a human outside the team will read. Also use when the user says "does this sound like AI", "de-slop this", "clean up this copy", "check the writing", "make this sound human". Strips the machine tells, enforces the banned list, and returns the text with a receipt. Never skipped on external-facing prose.

9 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

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

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When somebody asks whether something reads as machine-written

    line 14
  2. Distinctive:

    Delete or replace: unlock, leverage as a verb, supercharge, seamless, robust, holistic, synergy, game-changing, cutting-edge, …

    line 40
  3. Distinctive:

    Delete: "it's worth noting that", "it's important to remember", "arguably", "in many ways", "to some extent", and stacked qualifiers.

    line 53
  4. Distinctive:

    If a paragraph restates the previous paragraph in different words, delete it.

    line 57
  5. Distinctive:

    If the opening sentence could introduce any document on this topic, it is not an opening.

    line 61

“Common” is measured: kept by at least 6 in 10 of the 9 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

Writing Quality

A gate, not a suggestion. Any prose leaving the building passes through here first.

When to use

  • Automatically, on every piece of prose headed to an external reader
  • Any copy inside a deliverable, including copy presented as an option or a proposal
  • When somebody asks whether something reads as machine-written

Framing does not exempt text. Copy labeled "suggested", "draft", or "just an option" is still copy somebody will paste into an external surface, and it gets the same pass.

Inputs

  • Reads: workspace/agents/context/voice.md if it exists, otherwise the defaults below
  • Needs from user: nothing

Workflow

1. Strip the structural tells

These patterns are the clearest machine signatures in business prose, and they survive most editing passes because they read as confident.

PatternWhy it fails
"It's not X, it's Y"A rhetorical reveal that usually replaces an argument
"X didn't do Y, it revealed Z"The same move wearing a different coat
"The real X is..."Announces insight rather than delivering it
Rule-of-three lists everywhereReal thinking is rarely three parallel items
"In today's fast-paced world"Says nothing. Delete the sentence, never replace it
A closing paragraph restating the openingPadding. Stop at the last real point
Symmetrical two-part sentences on repeatRhythm that reads as generated after three of them

2. Strip the vocabulary

Delete or replace: unlock, leverage as a verb, supercharge, seamless, robust, holistic, synergy, game-changing, cutting-edge, best-in-class, elevate, empower, harness, navigate the landscape, at the end of the day, dive deep, double click, and any sentence containing "in an era of".

"Leverage" as a noun meaning negotiating power is fine. The verb is not.

3. Fix the punctuation habits

  • No em dashes. A period, a comma, or a rewrite. This is a hard rule
  • No semicolons in prose. Two sentences
  • Sparse parentheses. More than one aside per paragraph means the sentence structure is wrong
  • No exclamation marks in business prose

4. Cut the hedging

Delete: "it's worth noting that", "it's important to remember", "arguably", "in many ways", "to some extent", and stacked qualifiers. One qualifier is precision. Three is a lack of a position.

5. Cut the summary of what was just said

If a paragraph restates the previous paragraph in different words, delete it. Machine prose summarizes compulsively, and the summary is always the weaker version.

6. Read the first line honestly

If the opening sentence could introduce any document on this topic, it is not an opening. Start at the specific claim.

7. Strip invisible characters

Zero-width characters, bidirectional marks, tag characters, and exotic spaces travel invisibly through copy-paste and show up in other people's tooling. Scan and strip them as the last step, after wording is final.

Report the count.

Output

  • Returns the cleaned prose
  • Prints the receipt:
writing-quality: <n> structural tells, <n> vocabulary, <n> punctuation fixed
invisible-characters: <n> stripped | none

Rules & quality bar

  • This gate never gets skipped on external-facing prose, whatever the framing
  • No em dashes, no semicolons in prose. Hard rules
  • The receipt is mandatory. A missing receipt means the gate did not run, never that it passed silently
  • Wording is final before the invisible-character strip, so the strip changes nothing visible
  • Do not soften a claim while cleaning it. Cleaning is not hedging
  • Delete rather than replace filler. A cut sentence is better than a rewritten empty one

Related skills

  • Runs inside: artifact-eval as its first gate
  • Reads: the voice module from context-portfolio

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.