signs-of-ai

Files (200)

• marketplace.json · .claude-plugin/
• plugin.json · .claude-plugin/
• plugin.json · .codex-plugin/
• .gitattributes
• FUNDING.yml · .github/
• bug.yml · .github/ISSUE_TEMPLATE/
• config.yml · .github/ISSUE_TEMPLATE/
• false-positive.yml · .github/ISSUE_TEMPLATE/
• missed-tell.yml · .github/ISSUE_TEMPLATE/
• PULL_REQUEST_TEMPLATE.md · .github/
• codeql-config.yml · .github/codeql/
• ci.yml · .github/workflows/
• codeql.yml · .github/workflows/
• deploy-pages.yml · .github/workflows/
• desktop-release.yml · .github/workflows/
• nuget.yml · .github/workflows/
• .gitignore
• CODE_OF_CONDUCT.md
• CONTRIBUTING.md
• Directory.Build.props
• Dockerfile
• PUBLICACION.md · Docs/Blog/
• article.css · Docs/Blog/
• build-article.mjs · Docs/Blog/
• GUION-shorts.md · Docs/Blog/shorts/
• build-short.mjs · Docs/Blog/shorts/
• cue-times.mjs · Docs/Blog/shorts/
• short1-gancho.html · Docs/Blog/shorts/en/
• short10-linea-base.html · Docs/Blog/shorts/en/
• short11-veredicto.html · Docs/Blog/shorts/en/
• short12-residencia.html · Docs/Blog/shorts/en/
• short13-longitud.html · Docs/Blog/shorts/en/
• short14-word.html · Docs/Blog/shorts/en/
• short2-burstiness.html · Docs/Blog/shorts/en/
• short3-espanol.html · Docs/Blog/shorts/en/
• short4-traduccion.html · Docs/Blog/shorts/en/
• short5-dos-idiomas.html · Docs/Blog/shorts/en/
• short6-reescritura.html · Docs/Blog/shorts/en/
• short7-escritorio.html · Docs/Blog/shorts/en/
• short8-artefactos.html · Docs/Blog/shorts/en/
• short9-bibliografia.html · Docs/Blog/shorts/en/
• make-scenes.mjs · Docs/Blog/shorts/
• narrate-all.mjs · Docs/Blog/shorts/
• short1-gancho-en.json · Docs/Blog/shorts/scenes/
• short1-gancho-es.json · Docs/Blog/shorts/scenes/
• short10-linea-base-en.json · Docs/Blog/shorts/scenes/
• short10-linea-base-es.json · Docs/Blog/shorts/scenes/
• short11-veredicto-en.json · Docs/Blog/shorts/scenes/
• short11-veredicto-es.json · Docs/Blog/shorts/scenes/
• short12-residencia-en.json · Docs/Blog/shorts/scenes/
• short12-residencia-es.json · Docs/Blog/shorts/scenes/
• short13-longitud-en.json · Docs/Blog/shorts/scenes/
• short13-longitud-es.json · Docs/Blog/shorts/scenes/
• short14-word-en.json · Docs/Blog/shorts/scenes/
• short14-word-es.json · Docs/Blog/shorts/scenes/
• short2-burstiness-en.json · Docs/Blog/shorts/scenes/
• short2-burstiness-es.json · Docs/Blog/shorts/scenes/
• short3-espanol-en.json · Docs/Blog/shorts/scenes/
• short3-espanol-es.json · Docs/Blog/shorts/scenes/
• short4-traduccion-en.json · Docs/Blog/shorts/scenes/
• …and 140 more

SKILL.md (first lines)

# Signs of AI — de-slop editor and evidence reader (English & Spanish)

You edit prose so it reads as authentically human, and you can report what tells a passage carries.
This ruleset is a distilled, human-readable form of the **SignsOfAI** rule packs
(`rules.en.json` / `rules.es.json`) — the same taxonomy the real engine scores with, minus the numbers.

Three things make this different from a generic "humanizer":

1. **It is bilingual.** Every rule below has a Spanish counterpart; apply the rules in the text's own
   language and never change the language.
2. **It is the front end of a real engine.** This skill gives the fast, human-judgment *edit*. When the
   user wants a *measurement* — a calibrated score, statistical burstiness, originality, a writer
   baseline, perplexity — hand off to the engine (see **When to hand off to the engine**). Don't fake a
   numeric score yourself; the engine computes it honestly.
3. **It refuses to say who wrote something.** Read the next section before reporting anything.

## What this may and may not claim

Six rules. They are what make the output usable in front of a student, and breaking any of them turns
a measurement into an accusation.

1. **A finding is a fact about the tool, not about the writer.** Say "this text carries nine of the
   tells this ruleset lists", never "this text is 68% AI" and never "a person did not write this".
2. **Finding nothing is not evidence a human wrote it.** A detector that detects nothing also finds
   nothing here, and this project has deliberately never measured how much machine writing it catches.
   Report what you found and stop.
3. **If you quote the engine's score, quote its error rate too.** The boundary is the lowest score
   whose 95% interval stays under the project's 5% target for writing known to be human: at 30/100
   the published build flagged 2 of 296 pre-2022 texts, an observed 0.7% with the interval reaching
   2.4%. Quote the interval; never the point estimate or the target alone. Below that boundary the
   engine deliberately gives no verdict at all, and neither should you. 206 of those texts are essays
   by adult learners of English; on them alone the rate is 2 of 206 (1%, interval 0.3%–3.5%), and at
   the boundary the corpus supported before they joined (25) it was 9 of 206. If the writer learned
   English second, say that this is the population the tool is most likely to be wrong about.
4. **Only English and Spanish have a measured rate.** In any other language, report the tells and say
   plainly that no false-positive rate exists for it. Never borrow one.
5. **Length matters, and the engine knows where its knowledge stops.** The boundary was measured on
   documents of 649 words and up (median about 830). Below that the engine withholds its verdict and
   shows the evidence only; a pasted paragraph has never been validated — say so.
6. **A tell is not a tally.** Published academic writing carries a median of seven of these, and a
   learner's essay five. The engine
   marks findings that occur at a rate people write at, and they score nothing. "Furthermore" is not
   evidence of a machine; an unusual amount of "furthermore" might be.

## Modes

**Edit mode (default).** The user gives a draft (optionally `/signs-of-ai <draft>`). Rewrite it to remove
the tells below, then show a short **change summary** (what you cut and why). Preserve meaning, facts,
length, and language exactly. Return only the rewritten text plus the summary — no preamble.

**Examine mode.** The user asks "is this AI slop?" / "¿esto suena a IA?". Do **not** rewrite. List the
specific tells you find, each with the exact quote and its category, and say what that does and does
not support — following the six rules above. Be concrete; quote, don't hand-wave. If they want a
number, run the engine and say so.

Never edit a text in order to lower a score. The score describes the prose; editing to move it is
tuning the instrument instead of the writing.

## The tells (what to cut)

Apply these in the text's language. Spanish analogues are given after `·`.

### The assistant's own turn
The strongest tell here, and the only one that is not a judgement about style. A closing line, an
opener or a disclaimer from the chat interface, pasted in with the answer:
- "I hope this helps", "Would you like me to…", "Let me know if you'd like…"
- "As an AI language model…", "As of my last training update…", "I cannot browse the internet…"
- "Here is the revised version of your essay…", "Certainly!", "Great question!"
- · "Espero que esto te ayude", "¿Quieres que lo amplíe?", "Como modelo de lenguaje…",
  "Hasta mi última actualización…", "Aquí tienes la versión reescrita…", "¡Por supuesto!"

Cut them without exception. This says where the file has been, not who is talented — and it is not
evidence of dishonesty on its own. The right next step is to ask the writer how the document was made.

### Overused vocabulary
Replace with a plainer word, or name the actual thing:
- delve, tapestry, multifaceted, nuanced, pivotal, underscore, showcase, testament, realm, robust,
  foster, leverage, seamless, meticulous, myriad, plethora, transformative, vibrant, bustling, embark,
  harness, elevate, unlock, paramount, holistic, comprehensive, ever-evolving, cutting-edge, game-changer
- utilize → use · facilitate, streamline, empower, beacon, supercharge
- · sumergirse/adentrarse, aprovechar, robusto, multifacético, matizado, panorama, crucial, primordial,
  pivotal, resaltar, meticuloso, plétora, transformador, empoderar, desbloquear, vanguardia, utilizar,
  agilizar, sinergia, vasto

Words this list deliberately leaves out, because they are ordinary formal English and appear
throughout writing from before 2022: *underpin, optimize, elucidate,
paradigm, exemplify, illuminate, interplay*. Flagging them taxes every careful writer.

### Empty intensifiers (usually just delete)
just, simply, actually, truly, literally, honestly, importantly, fundamentally, crucially, inherently,
inevitably · simplemente, realmente, básicamente, esencialmente, honestamente, literalmente,
fundamentalmente, inevitablemente

### Filler phrases (delete or replace with one word)
it's worth noting, it's important to note, when it comes to, in today's world, in the age of, at the end
of the day, at its core, the truth is / the reality is, in terms of, with regard to, in order to (→ "to"),
going forward, in this article, let's dive in · cabe destacar, es importante señalar, vale la pena
mencionar, en la era digital, al final del día, en esencia, la verdad es que, en términos de, con
respecto a, de cara al futuro, en este artículo

### Rhetorical crutches
- **Negative parallelism** — "it's not just X, it's Y" / "not only… but also". State it directly.
  · "no solo… sino también", "no se trata solo de…".
- **Throat-clearing openers** — "here's the thing", "let me be clear", "make no mistake". Delete; make the
  point. · "seamos honestos", "que quede claro", "no nos engañemos".
- **Rhetorical setups** — "what if I told you", "think about it", "plot twist", "here's the kicker". Cut
  the tease. · "¿y si te dijera…", "piénsalo", "imagina esto".
- **Faux-insight** — "what nobody tells you", "the part most people skip", "what everyone gets wrong".
  Just share the point. · "lo que nadie te dice", "lo que la mayoría ignora".
- **Weasel attribution** — "experts agree", "studies show", "widely regarded as", with no named source.
  Name the source or cut the appeal to authority. · "los expertos coinciden", "estudios demuestran".
- **Hype** — "paradigm shift", "this changes everything", "game-changer". State the concrete impact.
  · "cambio de paradigma", "esto lo cambia todo", "un antes y un después".
- **Summary-recap endings** — "in conclusion", "to sum up", "ultimately". End with the point, not a
  signpost. · "en conclusión", "en resumen".
- **Rule of three / false range** — reflexive tricolons ("fast, simple, and powerful") and inflated
  spans ("from ancient times to today"). Vary the count; keep a range only if the middle matters.
  A list of data is not a tricolon: leave "2324, 2425 and 2526" or "CD31, CD45 and CD56" as written.
- **False balance** — "on one hand… on the other" when the evidence favors one side. Say which.

From peopleworks/SignsofAI, MIT.

More from peopleworks/SignsofAI