[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-81988":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":13,"openIssues":14,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":14,"stars7d":15,"stars30d":16,"stars90d":14,"forks30d":14,"starsTrendScore":17,"compositeScore":18,"rankGlobal":10,"rankLanguage":10,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":10,"pushedAt":10,"updatedAt":29,"readmeContent":30,"aiSummary":31,"trendingCount":14,"starSnapshotCount":14,"syncStatus":32,"lastSyncTime":33,"discoverSource":34},81988,"ai-detector-skill","lynote-ai\u002Fai-detector-skill","lynote-ai"," A free detector capable of identifying content generated by all advanced AI models.","https:\u002F\u002Flynote.ai\u002Fai-detector",null,"Python",226,4,0,21,165,3,2.1,"MIT License",false,"main",true,[24,25,26,27,28],"ai-agents","ai-detector","awesome-list","detector","skiils","2026-06-12 02:04:21","# Free AI Detector\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002Fhero.svg\" alt=\"AI Detector Skill hero\" width=\"100%\" \u002F>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flynote-ai\u002Fai-detector-skill\u002Factions\u002Fworkflows\u002Fci.yml\">\u003Cimg src=\"https:\u002F\u002Fgithub.com\u002Flynote-ai\u002Fai-detector-skill\u002Factions\u002Fworkflows\u002Fci.yml\u002Fbadge.svg\" alt=\"CI\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.9%2B-3776ab\" alt=\"Python 3.9+\" \u002F>\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-MIT-16a34a\" alt=\"MIT License\" \u002F>\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fnetwork-none-f59e0b\" alt=\"No Network\" \u002F>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  An explainable, cautious AI-generated text risk analyzer for coding agents and local workflows.\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\".\u002FREADME.zh-CN.md\">简体中文\u003C\u002Fa>\n  ·\n  \u003Ca href=\"#install\">Install\u003C\u002Fa>\n  ·\n  \u003Ca href=\"#evaluation\">Evaluation\u003C\u002Fa>\n  ·\n  \u003Ca href=\"#use-cases\">Use Cases\u003C\u002Fa>\n\u003C\u002Fp>\n\nThis project is intentionally modest. It estimates **AI-like signals**, not proof of authorship.\n\n## Why This Exists\n\nMost AI text detectors are either overconfident, opaque, or awkward to embed inside agent workflows.\n\n`ai-detector-skill` takes the opposite approach:\n\n- explainable weighted signals instead of hidden model claims\n- a local CLI and Python API that are easy to script\n- a short-text guardrail that refuses to overstate weak evidence\n- skill-ready packaging for Codex, Claude Code, and other repo-aware agents\n- reproducible benchmark and dataset evaluation scripts\n\nIf you want a **triage tool** that stays cautious and leaves room for human review, this repo is built for that.\n\n## Install\n\n```bash\npip install -e .\nai-detect examples\u002Fsample_ai_like.txt --json\n```\n\nOr bootstrap a local environment:\n\n```bash\nbash scripts\u002Fsetup.sh\n```\n\nExample output:\n\n```text\nConclusion: AI-like signals are present, but this medium-confidence score is a risk estimate rather than proof.\nScore: 84\u002F100\nConfidence: medium\nVerdict: high_ai_likelihood\nWords analyzed: 256\n```\n\n## What You Get\n\n- `score`: 0-100 AI-like writing risk estimate\n- `verdict`: `insufficient_text`, `low_ai_likelihood`, `mixed_or_uncertain`, or `high_ai_likelihood`\n- `confidence`: currently `low` or `medium`\n- `signals`: strongest weighted evidence signals\n- `caveats`: warnings that stay attached to the result\n- `next_steps`: practical follow-up actions when useful\n\nThis is designed to help an agent say:\n\n- \"AI-like signals are present.\"\n- \"The sample is too short for a meaningful estimate.\"\n- \"This should be reviewed against known writing samples.\"\n\nNot:\n\n- \"This was definitely written by AI.\"\n- \"The detector proves misconduct.\"\n\n## Usage\n\n### CLI\n\n```bash\nai-detect examples\u002Fsample_ai_like.txt\ncat essay.txt | ai-detect --json\npython scripts\u002Fdetect.py examples\u002Fsample_human_like.txt --json\n```\n\n### Python API\n\n```python\nfrom aidetect import analyze_text\n\ntext = open(\"essay.txt\", encoding=\"utf-8\").read()\nresult = analyze_text(text)\n\nprint(result.score, result.confidence, result.verdict)\nfor signal in result.strongest_signals():\n    print(signal.name, signal.note)\n```\n\n### Local Skill Install\n\n```bash\ncp -R ai-detector-skill \"$CODEX_HOME\u002Fskills\u002F\"\n```\n\nUse the root [SKILL.md](.\u002FSKILL.md) as the portable skill definition, and keep [AGENTS.md](.\u002FAGENTS.md) at the repo root for repo-aware agents.\n\n## Evaluation\n\nWe ran a reproducible evaluation on the public [HC3 dataset](https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002FHello-SimpleAI\u002FHC3), using the English `finance`, `medicine`, and `open_qa` subsets with the first 100 rows from each subset.\n\nSnapshot of the current detector on that slice:\n\n- Human mean score: `5.4`\n- AI mean score: `18.4`\n- Mean separation: `13.0` points\n- Human coverage: `0.427`\n- AI coverage: `0.920`\n- Covered accuracy at `score >= 45`: `0.317`\n\nWhat that means in practice:\n\n- the detector separates human and AI answers on average, but only weakly on HC3\n- the short-text guardrail is doing useful work, especially on shorter human answers\n- the current thresholds are conservative, which keeps false confidence down but also lowers recall\n- this tool works better as **triage + explanation** than as a stand-alone classifier\n\nSee the full report in [docs\u002FHC3_EVALUATION.md](.\u002Fdocs\u002FHC3_EVALUATION.md).\n\nReproduce it with:\n\n```bash\nmake eval-hc3\n```\n\n## Use Cases\n\n### Teacher Triage\n\nA teacher receives a polished 400-word reflection and wants a cautious signal before doing manual review.\n\nSuggested workflow:\n\n1. Run `ai-detect submission.txt --json`.\n2. Read the strongest signals and caveats.\n3. Compare the passage with known writing samples before making any judgment.\n\n### Editorial Review\n\nAn editor wants to spot formulaic product reviews or guest posts before spending time on manual edits.\n\nWhy it fits:\n\n- medium-length prose works better than short snippets\n- explainable signals help justify why a draft feels templated\n\n### Trust And Safety Queueing\n\nA moderation team wants to sort suspicious long-form posts into a manual review queue, not auto-remove them.\n\nWhy it fits:\n\n- the tool is conservative by design\n- it helps with prioritization more than enforcement\n\n### Internal Content QA\n\nA team compares human drafts and AI-assisted drafts to see where language starts sounding too generic.\n\nWhy it fits:\n\n- the score is useful as a relative signal across versions\n- strongest signals can guide rewriting\n\n## Not For\n\n- disciplinary decisions about a named student or employee\n- treating a single score as proof of cheating or fraud\n- very short samples under about 80 words\n- high-stakes authorship disputes without known-sample comparison\n\n## Project Structure\n\n```text\nai-detector-skill\u002F\n├── SKILL.md\n├── scripts\u002F\n│   ├── detect.py\n│   ├── setup.sh\n│   ├── benchmark.py\n│   └── evaluate_hc3.py\n├── references\u002F\n│   └── api-reference.md\n├── assets\u002F\n│   ├── hero.svg\n│   ├── score-bands.svg\n│   ├── workflow.svg\n│   └── templates\u002F\n│       └── report.md\n├── src\u002Faidetect\u002F\n├── tests\u002F\n├── AGENTS.md\n└── README.md\n```\n\n## Dev Commands\n\n```bash\nmake test\nmake demo\nmake benchmark\nmake eval-hc3\n```\n\n## CI\n\nGitHub Actions automatically runs:\n\n- `make test` on Python `3.9`, `3.11`, and `3.13`\n- `make benchmark` to regenerate the synthetic benchmark report\n- `make eval-hc3` to regenerate the HC3 evaluation report\n- upload of `docs\u002FBENCHMARK.md` and `docs\u002FHC3_EVALUATION.md` as workflow artifacts\n\n## Contributing\n\nSee [CONTRIBUTING.md](.\u002FCONTRIBUTING.md).\n\nGood contributions usually improve one of these:\n\n- clearer evidence signals\n- safer wording and UX\n- multilingual handling that stays explainable\n- reproducible evaluation coverage\n- agent integration ergonomics\n\n## License\n\nMIT\n","该项目是一个能够识别所有高级AI模型生成内容的免费检测器。它通过分析文本中的AI特征信号，为用户提供一个0到100的风险评分，并给出相应的判断结论如“高AI可能性”等，同时强调结果仅供参考而非确凿证据。采用Python 3.9+编写，支持本地CLI及Python API调用，便于集成进代码代理或本地工作流中。适用于需要对文档、代码等进行初步AI生成内容筛查的场景，特别是当需要保持谨慎态度并留有人工复核空间时。",2,"2026-06-11 04:07:23","CREATED_QUERY"]