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

writing-quality

Skill2 starsChanged 3 months agoMITView on GitHub

The anti-slop gate. Runs automatically inside every skill that produces copy (invite-motion, meeting-engine, event-followup, event-content, speaking-and-stage). Also use directly when the user says "does this sound like AI", "de-slop this", "clean up this invite", "make this sound human". Two modes: EDIT (rewrite, default) and DETECT (name the problems without rewriting).

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

6 rules. 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 6

    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:

    Symmetrical bullets: Every bullet the same length reads as generated

    line 41
  2. Distinctive:

    Complete sentences for executive audiences: No fragment runs, no "label: list" colon constructions

    line 57
  3. Distinctive:

    Plain practitioner vocabulary: Use the word the recipient would say out loud

    line 58
  4. Distinctive:

    No fabricated guest names: Never name an attendee who has not confirmed, and never without permission

    line 64
  5. Distinctive:

    Chatham House Rule is honored: If the room was off the record, no attributed quote leaves it

    line 67
  6. Distinctive:

    Run all six check groups. Note every hit with its location

    line 71

“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

An invitation that reads as automated to a senior executive is worse than no invitation, because it burns a name you cannot easily re-approach. This gate catches the tells before anything sends.

When to use

  • Automatically, inside every skill that produces copy. Not optional
  • Directly, on any draft
  • On copy someone else wrote that is about to go out with your executive's name on it

Modes

  • EDIT (default): rewrite, then list what changed and why
  • DETECT: name the problems and where they are, change nothing

The checks

1. Banned vocabulary

Each one is a rewrite, not a preference:

leverage, unlock, supercharge, seamless, robust, streamline, empower, elevate, harness, foster, facilitate, delve, embark, navigate (figurative), landscape (figurative), realm, tapestry, testament, journey (figurative), pivotal, crucial, paramount, meticulous, intricate, myriad, plethora, vibrant, transformative, revolutionary, game-changing, cutting-edge, groundbreaking, best-in-class, world-class, innovative, disruptive, holistic, synergy, paradigm shift, ecosystem (figurative), utilize, showcase, underscore, illuminate, reimagine, unveil, in today's fast-paced world, gone are the days, at the end of the day, that's where we come in

Plus the event-specific ones, which are their own dialect of slop:

exclusive, intimate (of a gathering), curated (when nothing was actually curated), immersive, unforgettable, one-of-a-kind, can't-miss, thought leaders (in prose), industry luminaries, the energy in the room, incredible turnout, we were thrilled to, join us as we, an evening of, a night to remember, save the date (as the entire message)

"Exclusive" and "intimate" are the two that most reliably mark an invitation as marketing. If a dinner is genuinely for twelve people, saying "a table of twelve" communicates the same thing and is believable.

2. Banned structures

  • "It's not X, it's Y" and every variant
  • "X didn't just do Y, it did Z"
  • "The real X." Reveal framing that reveals nothing
  • Rule of three everywhere. Three adjectives, three clauses, three bullets, repeatedly
  • Negative parallelism used as a rhythm rather than a distinction
  • Manufactured fragments. "Every. Single. Time."
  • The rhetorical question opener. "What if I told you..."
  • Symmetrical bullets. Every bullet the same length reads as generated

3. Punctuation and rhythm

  • Em dashes. The loudest tell in generated prose. Use a comma, a period, or parentheses
  • Semicolons in an email. Nobody writes a personal note with a semicolon in it
  • Every sentence the same length. Break the pattern

4. Substance checks

  • Vague attribution. "Studies show", "industry leaders agree". Name it or cut it
  • Unsupported superlatives
  • Padding. "It's important to note that", "when it comes to", "in order to"
  • Hedging stacks
  • Restating the prompt

5. Register checks for event copy

  • Contractions in first-person writing. A contraction-free note from a named human reads as machine formality
  • Complete sentences for executive audiences. No fragment runs, no "label: list" colon constructions
  • Plain practitioner vocabulary. Use the word the recipient would say out loud
  • The forward test. Would the sender be comfortable if the recipient forwarded this to a colleague?
  • The single-recipient test. Could this sentence have been sent to a thousand people? If yes, it is not a personal invitation

6. Claim integrity

  • No invented statistics. Hard fail
  • No fabricated guest names. Never name an attendee who has not confirmed, and never without permission
  • No claimed familiarity that does not exist. No "as we discussed" with someone you have not spoken to, no invented mutual connection. Executives check
  • Quotes are verbatim or they are paraphrase without quote marks
  • Chatham House Rule is honored. If the room was off the record, no attributed quote leaves it

Workflow

  1. Read the draft
  2. Run all six check groups. Note every hit with its location
  3. In EDIT mode, rewrite. Preserve the argument and the sender's voice. Do not over-scrub. Removing every rough edge produces flat, safe copy, which for an invitation is its own failure
  4. Re-read for the forward test and the single-recipient test
  5. Return the clean version plus a short change list

Output

  • EDIT: the rewritten copy, then the change list
  • DETECT: findings with locations and severity, no rewrite

Rules & quality bar

  • Invented numbers and fabricated guest names are hard fails
  • The single-recipient test is the one that matters most for invitations
  • Do not over-scrub. Human, not neutral
  • Preserve the sender's voice. Fix the tells, not the personality
  • Report what changed

Related skills

  • Runs inside: invite-motion, meeting-engine, event-followup, event-content, speaking-and-stage
  • Escalates to: artifact-eval when a scored judgment is needed

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.