ai-check
Use when someone asks "does this sound AI?", "check if this is AI-written", "what gives this away as AI", "run ai-check on this", or "score this text". Also use when reviewing a draft for AI tells before publishing, or when a piece of text reads as suspiciously polished, generic, or pattern-y and the user wants a forensic breakdown of why.
npx skills add harshaneel/humanize --skill ai-check --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Forensic analysis of text for AI-generation signals. Grounded in the published detection literature. The output is a structured report, not a vague judgment. Every fired signal cites evidence.
How it works
The skill defines nine signal categories (A–I). For each category, it assigns a score from 0 to 3 based on detected signals such as perplexity, burstiness, hedge density, structural tells, specificity deficit, transition word usage, punctuation fingerprints, voice/register, and rhetorical scaffolding. It specifies how to log signals, quote exact phrases that fired, and apply a double-counting policy where a single phrase can fire multiple signals only if genuinely diagnostic for different signals. It provides detailed flags, criteria, and examples for each category and instructs to report exact quotes, counts, and itemized observations. It also imposes a total score cap of 27 and requires citing evidence for every signal.
When to use it
Use when someone asks if text sounds AI-generated, to check if text is AI-written, to identify what gives this away as AI, to run an ai-check on text, or to score text. Also applicable when reviewing a draft for AI tells before publishing or when text appears suspiciously polished or pattern-y and a forensic breakdown is desired.
What it can touch
The skill declares tools: claude-code, codex. It instructs using these tools to perform the forensic analysis and cite evidence accordingly.
Caveats
The skill references detection literature and provides structured signals and scoring; it does not guarantee a human/machine authorship verdict. The approach emphasizes citing evidence for every fired signal and adheres to a predefined scoring scheme. It does not include outcomes beyond the structured report as described.
# AI-Check Skill Forensic analysis of text for AI-generation signals. Grounded in the published detection literature (Wu et al. 2025, Mitchell et al. 2023, Kujur 2025, AAAI 2025 shared task). The output is a structured report, not a vague judgment. Every fired signal cites evidence. --- ## The nine signal categories Score each category 0–3: - 0 = No signal detected (human-consistent) - 1 = Weak signal (possible AI, could be human) - 2 = Moderate signal (likely AI pattern) - 3 = Strong signal (near-certain AI pattern) **Severity-to-score mapping (use for every category):** | Evidence in category | Score | |---|---| | No flagged instances | 0 | | One weak instance, or vague unease without a specific quote | 1 | | One moderate instance, or two or more weak instances | 2 | | One strong instance, or two or more moderate instances, or four or more weak instances | 3 | **Double-counting policy:** a single phrase can fire at most two distinct signals when the phrase is genuinely diagnostic for both. Example: "it is important to note that" is both Signal A (banned vocabulary) and Signal C (institutional hedge). Log it under both, but the same phrase cannot count as two separate weak instanc
- The nine signal categories
- Signal A: Perplexity (word predictability)
- Signal B: Burstiness deficit (sentence uniformity)
- Signal C: Hedge density
- Signal D: Structural tells
- Signal E: Specificity deficit
- Signal F: Transition word fingerprint
- Signal G: Punctuation fingerprint
- Signal H: Voice and register
- Signal I: Rhetorical scaffolding
- Mixed-authorship overlay (estimate how much AI editing)
- Output format
- Scoring thresholds
- Calibration notes
What does the ai-check skill do?
Use when someone asks "does this sound AI?", "check if this is AI-written", "what gives this away as AI", "run ai-check on this", or "score this text". Also use when reviewing a draft for AI tells before publishing, or when a piece of text reads as suspiciously polished, generic, or pattern-y and the user wants a forensic breakdown of why.
How do I install it?
Run `npx skills add harshaneel/humanize --skill ai-check --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From harshaneel/humanize, a repository with 330 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.