AI Text Detector skill
A skill finds AI-like writing signals with a local analyzer and explains why its score does not establish who wrote the text.
- 663 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skill can save supplied text to a local file; if a cloud agent reads it, the model provider receives it.
Install
- Works in
Manual install
pip install -e .Installs the local analyzer from source. Requires Python 3.9 or later; add SKILL.md to the agent separately. This is the README’s first command; we did not run it.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill can save supplied text to a local file; if a cloud agent reads it, the model provider receives it.
This is third-party code. Review the repository files before installing.
What it does
AI Text Detector checks text for language signals and shows what affected the score. Its local CLI returns a score, confidence level and explanations. The skill helps an agent read the report and select passages for human review. If the passage is too short, the agent asks for a longer sample. For consequential decisions, the instructions recommend comparing it with known writing samples. The repository also includes code for training a separate model; the commercial detector is not part of the local installation.
Who it is for. For editors who want to locate formulaic wording and understand the score before reviewing a text by hand.
Good fit when
- Identify wording that looks formulaic.
- Return structured JSON in a local editorial workflow.
- Compare reported signals with the source passage.
- State the limits of a short sample before human review.
Not a fit when
- You need proof of misconduct or grounds to punish a named person.
- You need to establish the origin of one short sentence.
- You expect a trained production model in the local installation.
Example request
Review this text locally, showing signals and limitations without making claims about the author’s identity.Limitations
The score reflects language signals and does not prove authorship; accuracy for a new genre or language is not guaranteed. For passages shorter than roughly 80 words, the skill asks for more text because the sample gives little basis for a conclusion. The separate model’s training code ships without weights or datasets, so installation does not provide the commercial detector. The CLI works without networking, but text given to a cloud-model agent reaches that model’s provider.
How to disable. Remove the agent skill and stop invoking `ai-detect`. Manage saved source texts and JSON reports separately.
FAQ
Does the local analyzer send text to an API?
No. The lightweight CLI is designed without network access; Lynote’s hosted API is a separate service.
Which fields does it return?
Score, confidence, verdict, signals and caveats. The skill validates these before replying.
Can it run without the installed command?
Yes. Use `python -m aidetect.cli path/to/text.txt --json` or the repository helper.
What does the separate model’s training code include?
It prepares data, trains the model and evaluates its scores. Trained weights and datasets are not included.
Can it review an article or email?
Yes. Both are documented prose types, with results remaining signals for an editor.
Discussion
No questions or reviews yet.
Related
SkillsTwo skills for English prose: humanize rewrites a model draft in a natural voice, and ai-check flags machine-writing tells
Avoid AI Writing
avoid-ai-writing
A skill that finds typical machine-writing patterns in text and rewrites them, with detect-only and edit-in-place modes
Humanizer (brandonwise)
humanizer
A skill and Node.js CLI that score English text for AI-writing patterns and statistics and suggest fixes on a 0 to 100 scale
A skill that finds AI writing tells, rewrites the text in a chosen voice, and optionally scores it from 0 to 100
A skill that edits English drafts: strips AI boilerplate, keeps the author's voice, and can also just list what it found without rewriting
A Russian text editing skill: removes bureaucratese, calques and AI mannerisms without touching facts, with a scanner to measure the edit
README in short
The skill requires an explanation of the score, signals and caveats. Tests on prepared samples do not establish accuracy for every text an editor may encounter.
README badge
Original SKILL.md
--- name: ai-text-detector-github description: Detect whether a passage shows AI-like writing signals and return an explainable risk estimate with confidence, caveats, and next steps. --- # AI Text Detector Skill ## Overview This skill wraps the local `ai-detector-skill` analyzer into a reusable open source Codex-style skill. It is designed for: - essays - emails - articles - reviews - forum posts - other prose where a user asks whether the writing may be AI-generated This skill returns a risk estimate, not proof of authorship. ## Trigger Conditions Use this skill when: - the user asks "is this AI-written?" - the user asks to detect AI-generated text - the user pastes a passage and asks whether it sounds machine-written - the user wants a cautious AI-likeness review for a document or message ## Required Behavior - Never present the score as proof. - Never accuse a named person of cheating, fraud, or misconduct. - Ask for a longer sample when the text is under about 80 words. - Prefer "AI-like signals are present" over "This was written by AI". - For high-stakes contexts, recommend human review and comparison with known writing samples. ## Execution Flow 1. Receive the candidate text and check length. 2. If the sample is under about 80 words, explain that the detector will be noisy and ask for a longer sample when possible. 3. Save the text to a file or pipe it through stdin. 4. Run the local detector: