ai-written-check
Skill0 starsChanged 3 months agoMITView on GitHub
Detects the mechanical tells that make professional prose read as AI-generated — em dashes, triple-list anaphora, "not X but Y" antithetical pairs, spatial and economic metaphors for abstract ideas, "architect" as a verb, engineered sentence cadence, mid-paragraph register swerves, builder/startup slang, sincerity adverbs ("honestly", "quietly"), and unfalsifiable universal claims ("everyone", "across the board"). Use when reviewing or editing a resume, cover letter, portfolio page, case study, proposal, bio, or any human-facing copy, or when the user asks for an "AI-written check", "AI tells", says copy "sounds like AI", or asks to make writing sound less generated.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- cabbagecachekid/red-penword for word, first seen Aug 16, 2026
Other versions1 project uses a version of this.
Word for word · 1
- cabbagecachekid/red-pen★ 0100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository7 other files it gives its agents.
- copyright-creative-workSkill
- cringe-checkSkill
- full-reviewSkill
- journey-mapSkill
- line-editSkill
- using-neon-jetpackSkill
- ux-web-design-reviewSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
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
- Distinctive: line 14
Read the copy once for sense.
- Distinctive: line 19
Never flag without a rewrite.
≈ “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.
Files
AI-Written Check
Scan human-facing prose for the mechanical patterns that signal machine authorship, and propose a concrete rewrite for each. This is a prose check — it looks at sentence mechanics and word choice, not at tone or positioning (that is the cringe-check skill).
What this is for. This is not about hiding that AI was involved. It assumes the healthy case: you're collaborating with Claude to refine, expand, or draft your own writing. The job is to keep the result aligned with your intent and your voice — so the words still sound like you wrote them, not like a model smoothed them into generic copy. None of these skills are for making AI do the writing for you.
How to run it
- Read the copy once for sense.
- Walk the checklist below. For every hit, output: the rule name, the offending text quoted, and a specific rewrite — not just "this sounds AI."
- Respect the thresholds. Some patterns are fine once and only become a tell when repeated. Count occurrences before flagging.
- End with a one-line verdict: clean, light touch-up, or heavy.
Never flag without a rewrite. The point is to fix the prose, not to label it.
The checklist
| Tell | Threshold | Fix |
| Em dash (—) | The #1 tell. Flag almost every one in body copy. | Period, colon (sparingly), or a shorter sentence. Acceptable only as a structural separator in a header. |
| Triple-list anaphora ("about money, about trust, about control") | 1 per page is fine; 2+ is a tell. | Cut to a single clause or break into separate sentences. |
| "Not X, but Y" / "X, not Y" antithetical pairs | 1 per piece is fine; more is a tell. | State the point directly without the contrast scaffold. |
| Spatial metaphors for abstract concepts ("corners of their financial lives", "the territory of taxes", "the current frontier") | Flag each. | Use the concrete noun the metaphor is standing in for. |
| Economic/cost metaphors for non-economic things ("make rigor the cheap option", "the time tax", "the heavy lift") | Flag each. | Say the literal thing. |
| "Architect" as a verb | Flag each. | "Design" or "build." |
| "Across" overuse ("across the funnel", "across products", "across teams") | Flag clusters. | Name the specific thing, or drop the word. |
| "From X to Y to Z" career/idea arcs | Flag when a plainer sentence works. | Rewrite as a direct sentence. |
| Engineered sentence cadence (deliberate alternation of very short and very long sentences) | Flag when the rhythm feels manufactured. | Let sentence length follow meaning, not a beat. |
| Voice-register swerve | Flag any paragraph that opens concrete-and-human and pivots to abstract industry-speak. | Stay in one register; the swerve is the tell. |
| Builder/startup subculture slang ("vibe coding", "shipping", "hot takes", "cracked at") | Flag each in formal copy. | A neutral substance-equivalent term. |
| Unfalsifiable universal claims ("everyone", "all", "every", "universally", "across the board") | Question each. | Scope it to the specific people/context actually observed. "Affects everyone" is itself an AI-pitch tell. |
| Sincerity adverbs ("honestly", "quietly", "genuinely", "truly", "frankly" as dressing) | Flag each. | Delete the adverb; if the sentence loses nothing, it was a tell. |
| Generic placeholder / template copy (fill-in-the-blank headers, "Welcome to my portfolio", "I'm passionate about X", boilerplate section intros that could belong to anyone) | Flag each. | Replace with something specific and true to the author — their actual words, project, or point of view. Filler that "isn't the vibe" is a tell that the line was generated, not authored. |
Why these matter
These patterns are how a reader's pattern-matcher flips to "a model wrote this." Most are individually harmless — the damage is density. One em dash is invisible; six is a signature. Apply the thresholds rather than nuking every instance, or the rewrite reads as over-corrected.
For the full catalog with worked before/after examples and edge cases, see references/TELLS.md.
What this skill does NOT do
- It does not judge tone, arrogance, or positioning — use
cringe-check. - It does not verify facts or claims — that belongs to a full review (
full-review). - It does not rewrite wholesale. It flags and proposes; the author decides.
Register extensions
references/TELLS.md is the core catalog for every register. If a register is resolved (see full-review: .red-pen/profile.md → ask once → default job-seeker), also apply the tells under its ## AI-tell extensions — for publication that adds engineered contrast constructions, stock AI vocabulary, false-balance closers, symmetrical paragraph engineering, and restating summary paragraphs. Report core and extension hits under one heading; note which register supplied each extension.
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
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.