Skip to content
Back to search

SkillNet

Create, Evaluate, and Connect AI Skills

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/zjunlp--skillnet/badge.svg)](/listings/zjunlp--skillnet)

HTML

<a href="/listings/zjunlp--skillnet"><img src="/api/listings/zjunlp--skillnet/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Requires Python 3.10 or newer.

pip install skillnet-ai

Install optional dependencies for analysis and routing:

pip install "skillnet-ai[graph]"         # scenario analysis
pip install "skillnet-ai[graph,claude]"  # analysis and routing via Claude
pip install "skillnet-ai[graph,codex]"   # analysis and routing via Codex

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 490 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <div align="center">
  2. <a href="http://skillnet.openkg.cn/">
  3. <img src="assets/brand/skillnet.png" width="190" alt="SkillNet Logo">
  4. </a>
  5. # SkillNet
  6. **Open infrastructure for discovering, evaluating, analyzing, and routing reusable AI agent skills.**
  7. [![PyPI version](https://badge.fury.io/py/skillnet-ai.svg)](https://pypi.org/project/skillnet-ai/)
  8. [![GitHub stars](https://img.shields.io/github/stars/zjunlp/SkillNet?style=social)](https://github.com/zjunlp/SkillNet)
  9. [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
  10. [![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
  11. [![arXiv](https://img.shields.io/badge/arXiv-b5212f.svg?logo=arxiv)](https://arxiv.org/abs/2603.04448)
  12. [![Hugging Face](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-FFD21E)](https://huggingface.co/blog/xzwnlp/skillnet)
  13. [![Website](https://img.shields.io/badge/Website-skillnet.openkg.cn-0078D4.svg)](http://skillnet.openkg.cn/)
  14. [![On StackMap](https://img.shields.io/endpoint?url=https%3A%2F%2Fstackmap.shipwithai.xyz%2Fapi%2Fbadge%2Fskillnet.json)](https://stackmap.shipwithai.xyz/repos/zjunlp/skillnet?utm_source=badge)
  15. [Website](http://skillnet.openkg.cn/) · [Python SDK](./skillnet-ai) · [Local UI](#local-browser-interface) · [Examples](./examples) · [Experiments](./experiments) · [Paper](https://arxiv.org/abs/2603.04448)
  16. </div>
  17. <div align="center">
  18. <video src="https://github.com/user-attachments/assets/a18cf8a5-4842-4774-be75-5481d08e1ba3" controls="controls" width="800" preload="metadata"></video>
  19. </div>
  20. ---
  21. SkillNet provides unified infrastructure for the agent skill lifecycle:
  22. - **Discovery:** search a public skill library by keyword or semantic intent.
  23. - **Installation:** download skill folders from GitHub into local agent workspaces.
  24. - **Creation:** generate structured skills from repositories, documents, prompts, or execution traces.
  25. - **Evaluation:** assess skills for safety, completeness, executability, maintainability, and cost awareness.
  26. - **Analysis:** extract capabilities and usage scenarios, and infer relationships between local skills.
  27. - **Routing:** select skills for a task from your local library, with selection reasons.
  28. <div align="center">
  29. ![SkillNet overview: skill categories, relationships, and evaluation dimensions](assets/media/skillnet-overview.gif)
  30. </div>
  31. ---
  32. ## News
  33. - **[2026-09-24]** [**skillnet-ai 0.1.2**](https://pypi.org/project/skillnet-ai/0.1.2/)
  34. adds a [local browser interface](#local-browser-interface) for exploring skill
  35. folders, analyzed relationships, and source evidence. Launch it with `skillnet ui`.
  36. - **[2026-08-20]** The updated [**SkillNet report**](https://arxiv.org/abs/2603.04448) presents
  37. **SkillNet-Gym**, with executable benchmarks for skill construction, retrieval, and composition,
  38. and **SkillNet-Fabric**, which routes tasks through a Wiki built for each task.
  39. - **[2026-07-11]** The [**SkillNet library**](http://skillnet.openkg.cn/) now indexes
  40. **500K+ GitHub skills**, with improved deduplication and broader coverage of scientific research
  41. and data analysis. This update also adds local scenario graphs and orchestration.
  42. - **[2026-03-26]** [**JiuwenClaw**](./examples/JiuwenClaw/README.md) integrates SkillNet as a
  43. built-in skill marketplace, bringing skill discovery and installation into its agent workflow.
  44. - **[2026-03-12]** The [**SkillNet MCP server**](https://github.com/CycleChain/skillnet-mcp),
  45. maintained by [CycleChain](https://github.com/CycleChain), makes SkillNet tools available to
  46. MCP-compatible agents.
  47. - **[2026-03-04]** [**SkillNet: Create, Evaluate, and Connect AI Skills**](https://arxiv.org/abs/2603.04448)
  48. is now available on arXiv, describing the project's approach to reusable agent skills.
  49. - **[2026-02-23]** [**OpenClaw**](https://github.com/openclaw/openclaw) now includes SkillNet
  50. as a built-in skill for discovering and reusing agent capabilities.
  51. ---
  52. ## Web Platform
  53. Explore the public skill library on [skillnet.openkg.cn](http://skillnet.openkg.cn/).
  54. - **Find skills:** search the [skill library](http://skillnet.openkg.cn/resources) by keyword or semantic intent, and filter by category.
  55. - **Review skills:** inspect descriptions, evaluation ratings, and GitHub sources, with download links for reuse.
  56. - **Explore collections:** browse [curated skill collections](http://skillnet.openkg.cn/package) for specific domains and tasks, and inspect the relationships between their skills.
  57. The site also introduces [SkillNet-Gym](http://skillnet.openkg.cn/skillgym) for skill graph construction and lifecycle benchmarking, and [SkillNet-Fabric](http://skillnet.openkg.cn/skillfabric) for Wiki-based skill routing, including research results and a guided routing demo.
  58. <div align="center">
  59. <video src="https://github.com/user-attachments/assets/9f9d35b0-36fd-4d7d-a072-39afa380b241" controls="controls" width="800" preload="metadata"></video>
  60. </div>
  61. ---
  62. ## Local Browser Interface
  63. The [browser interface](skillnet-ai/src/skillnet_ai/web/ui/README.md) is part of the
  64. Python package. Browse your own skill folders, inspect saved analysis graphs,
  65. and follow relationship evidence to the original source lines.
  66. Available in **skillnet-ai 0.1.2+**. Install and launch it with:
  67. ```bash
  68. pip install "skillnet-ai[ui]>=0.1.2"
  69. skillnet ui --skills-dir "/absolute/path/to/skills"
  70. ```
  71. Open `http://127.0.0.1:8765`. Choose **Current skill files** to read the sources or
  72. **Analysis results** to explore saved `compose_with` and `similar_to` relationships.
  73. You can also import a `graph.json` generated by `skillnet analyze`.
  74. Packaged releases include the website; users do not need Node.js or model keys
  75. to browse. Generate analysis separately with the SDK or CLI, then view the same
  76. results here. The interface starts with an empty personal library.
  77. ---
  78. ## Python SDK
  79. ### Install
  80. Requires Python 3.10 or newer.
  81. ```bash
  82. pip install skillnet-ai
  83. ```
  84. Install optional dependencies for analysis and routing:
  85. ```bash
  86. pip install "skillnet-ai[graph]" # scenario analysis
  87. pip install "skillnet-ai[graph,claude]" # analysis and routing via Claude
  88. pip install "skillnet-ai[graph,codex]" # analysis and routing via Codex
  89. pip install "skillnet-ai[graph,dsh]" # analysis and routing via DeepSeek Harness
  90. ```
  91. ### Initialize
  92. ```python
  93. from skillnet_ai import SkillNetClient
  94. client = SkillNetClient()
  95. ```
  96. The client reads environment variables and saved settings. See
  97. [Configuration](#configuration) to set up model endpoints or GitHub authentication.
  98. ### Search
  99. Search the hosted SkillNet catalog by keyword or semantic intent. Keyword search
  100. uses BM25 and ranks results by relevance by default; vector search uses the
  101. search service's embeddings.
  102. The SDK and CLI default to keyword relevance starting with `skillnet-ai` 0.1.4.
  103. Upgrade with `pip install --upgrade skillnet-ai`; use `sort_by="stars"` or
  104. `--sort-by stars` for explicit star-based sorting.
  105. ```python
  106. results = client.search(
  107. q="analyze financial PDF reports",
  108. mode="vector",
  109. threshold=0.85,
  110. limit=10,
  111. )
  112. for skill in results:
  113. print(skill.skill_name, skill.stars, skill.skill_url)
  114. ```
  115. ### Download
  116. Download a GitHub skill folder and its resources, check its structure, and return
  117. the installed path. Use `overwrite=True` to replace an existing folder.
  118. ```python
  119. local_path = client.download(
  120. url="https://github.com/anthropics/skills/tree/main/skills/pdf",
  121. target_dir="./my_skills",
  122. )
  123. print(local_path)
  124. ```
  125. ### Create
  126. Generate skill packages from a prompt, repository, document, or execution trace.
  127. Choose one source per call; the result is a list of generated directory paths.
  128. ```python
  129. paths = client.create(
  130. prompt=(
  131. "Create a csv-quality-checker skill that checks CSV files for missing "
  132. "values and duplicate rows without modifying the input."
  133. ),
  134. output_dir="./my_skills",
  135. )
  136. print(paths)
  137. client.create(
  138. github_url="https://github.com/zjunlp/DeepKE",
  139. output_dir="./my_skills",
  140. )
  141. client.create(
  142. office_file="./guide.pdf",
  143. output_dir="./my_skills",
  144. )
  145. ```
  146. ### Evaluate
  147. Assess a local skill or GitHub skill URL across five quality dimensions. Each
  148. dimension contains a `level` (`Good`, `Average`, or `Poor`) and a `reason`.
  149. Evaluation reviews the skill's instructions and supporting files without executing
  150. its scripts.
  151. ```python
  152. report = client.evaluate("./my_skills/pdf")
  153. print(report["safety"]["level"], report["safety"]["reason"])
  154. print(report["maintainability"]["level"], report["maintainability"]["reason"])
  155. ```
  156. ### Analyze
  157. Extract scenarios and capabilities from local skills, then build a reusable
  158. relationship graph, retrieval index and Wiki for routing.
  159. ```python
  160. analysis = client.analyze("./my_skills", output_dir="./skillnet_index")
  161. print(analysis.index_dir)
  162. print(analysis.skill_count, analysis.relation_counts)
  163. ```
  164. The graph has two relations: directed `compose_with` for an output or state from
  165. one skill that supports another, and undirected `similar_to` for comparable
  166. capabilities. Both are tied to specific scenarios.
  167. `analyze` reads direct child folders containing `SKILL.md`. Its default output is
  168. `<skills_dir>/.skillnet`; pass `output_dir` to choose another location. Reanalyze
  169. after editing skills to update the snapshot used by routing.
  170. See [analysis options and index details](skillnet-ai/README.md#analyze).
  171. ### Route
  172. Select up to `k` skills for a task using an index created by `analyze`. Hybrid
  173. search and graph expansion identify candidates; a Claude, Codex or DeepSeek Harness SDK
  174. compares them in a task Wiki containing skill profiles, relationships, and source text.
  175. ```python
  176. result = client.route(
  177. "Extract tables from my PDF report and check the resulting CSV "
  178. "for missing values and duplicate rows.",
  179. index_dir="./skillnet_index",
  180. k=5,
  181. )
  182. for skill in result.skills:
  183. print(skill.skill_id, skill.name, skill.path, skill.reason)
  184. ```
  185. Relations guide candidate exploration; the task and each skill's constraints
  186. determine the final selection.
  187. `route` returns `skills` (`skill_id`, `name`, `path`, `reason`) and available SDK
  188. `usage`. It selects up to `k` skills, possibly none. The selected set may cover
  189. only part of a task, and its order does not specify execution order.
  190. See [routing options](skillnet-ai/README.md#route) and
  191. [endpoint configuration](skillnet-ai/README.md#analysis-and-routing-endpoints).
  192. ## CLI
  193. The CLI exposes the same six operations. Use `python -m skillnet_ai` wherever
  194. the `skillnet` command is unavailable on your PATH.
  195. | Command | What it does | Example |
  196. | :-- | :-- | :-- |
  197. | `search` | Search SkillNet | `skillnet search "pdf" --mode vector` |
  198. | `download` | Install a skill | `skillnet download <url> -d ./my_skills` |
  199. | `create` | Create a skill package | `skillnet create --prompt "A skill for table extraction"` |
  200. | `evaluate` | Evaluate a local or remote skill | `skillnet evaluate ./my_skill` |
  201. | `analyze` | Build a local scenario graph and index | `skillnet analyze ./my_skills --output-dir ./skillnet_index` |
  202. | `route` | Select local skills for a task | `skillnet route "analyze my CSV" --index-dir ./skillnet_index` |
  203. | `ui` | Browse local skills and saved analysis | `skillnet ui --skills-dir /absolute/path/to/skills` |
  204. Use `skillnet <command> --help` for full options, or see the
  205. [CLI examples](skillnet-ai/README.md#cli). Add `--json` for a structured
  206. `{ok, data, error}` response on stdout; logs go to stderr.
  207. `create` checks the generated package structure; add `--evaluate` to request a
  208. model assessment as well. Use `skillnet validate <skill_dir>` for a local structure
  209. check without a model call.
  210. ## Configuration
  211. Search and public downloads need no API key. Create, evaluate, and analyze use an
  212. OpenAI-compatible Chat Completions endpoint. Analyze also needs an embedding
  213. endpoint; route uses that same embedding endpoint and a separately configured
  214. Claude, Codex or DeepSeek Harness SDK.
  215. Settings resolve in this order: explicit arguments → environment variables →
  216. user configuration → defaults. Load any `.env` file into your environment before
  217. calling the SDK; it does not load one automatically.
  218. | Variable | Purpose | Default |
  219. | :-- | :-- | :-- |
  220. | `API_KEY` | `create`, `evaluate`, `analyze` | unset |
  221. | `BASE_URL` | Chat Completions endpoint for create, evaluate and analyze | `https://api.openai.com/v1` |
  222. | `SKILLNET_MODEL` | Model for create, evaluate, and analyze | `gpt-4o` |
  223. | `GITHUB_TOKEN` | Private repos or higher GitHub rate limits | unset |
  224. | `GITHUB_MIRROR` | Public download fallback; disabled with GitHub authentication | unset |
  225. | `SKILLNET_API_URL` | Hosted search service URL | `http://api-skillnet.openkg.cn` |
  226. | `EMBEDDING_API_KEY` | `analyze` and `route` | unset |
  227. | `EMBEDDING_BASE_URL` | `analyze` and `route` | unset |
  228. | `EMBEDDING_MODEL` | `analyze` and `route` | unset |
  229. | `SKILLNET_EXPLORER_BACKEND` | `route`: claude, codex or dsh | `claude` |
  230. | `SKILLNET_EXPLORER_API_KEY` | `route` SDK credential | unset |
  231. | `SKILLNET_EXPLORER_BASE_URL` | `route` SDK-compatible base URL | unset |
  232. | `SKILLNET_EXPLORER_MODEL` | `route` SDK model | unset |
  233. Linux and macOS:
  234. ```bash
  235. export API_KEY="your-api-key"
  236. export BASE_URL="https://api.openai.com/v1"
  237. export SKILLNET_MODEL="gpt-4o"
  238. ```
  239. Windows PowerShell:
  240. ```powershell
  241. $env:API_KEY = "your-api-key"
  242. $env:BASE_URL = "https://api.openai.com/v1"
  243. $env:SKILLNET_MODEL = "gpt-4o"
  244. ```
  245. Analysis and routing require additional settings from the table above. See
  246. [endpoint configuration](skillnet-ai/README.md#analysis-and-routing-endpoints)
  247. for complete examples. `skillnet configure --interactive` saves settings locally;
  248. `skillnet doctor --json` checks configuration and optional dependencies.
  249. ---
  250. ## REST API
  251. The SkillNet search API is public and requires no authentication.
  252. ```bash
  253. curl "http://api-skillnet.openkg.cn/v1/search?q=pdf&limit=5"
  254. curl "http://api-skillnet.openkg.cn/v1/search?q=reading%20charts&mode=vector&threshold=0.8"
  255. ```
  256. <details>
  257. <summary><b>Search API parameters</b></summary>
  258. **Endpoint:** `GET http://api-skillnet.openkg.cn/v1/search`
  259. | Parameter | Type | Default | Description |
  260. | :-- | :-- | :-- | :-- |
  261. | `q` | string | required | Search query, keywords or natural language |
  262. | `mode` | string | `keyword` | `keyword` or `vector` |
  263. | `category` | string | unset | Filter by category |
  264. | `limit` | int | `10` | Results per page, max 50 |
  265. | `page` | int | `1` | Page number, keyword mode only |
  266. | `min_stars` | int | `0` | Minimum star count, keyword mode only |
  267. | `sort_by` | string | `relevance` | `relevance`, `stars` or `recent`, descending; keyword mode only |
  268. | `threshold` | float | `0.8` | Similarity threshold, vector mode only |
  269. </details>
  270. ---
  271. ## Use SkillNet Inside Agents
  272. Install the [SkillNet skill](skills/skillnet/) to give your agent access to all
  273. six operations. See [installation and configuration](skills/skillnet/references/setup.md)
  274. and [agent setup](skills/skillnet/references/platforms.md) for instructions.
  275. The demo below shows Claude Code using the SkillNet skill.
  276. <div align="center">
  277. <video src="https://github.com/user-attachments/assets/ae6020d9-6846-4672-84ce-fa9c8057e92b" controls="controls" width="800" preload="metadata"></video>
  278. </div>
  279. ### Model Context Protocol
  280. The community [SkillNet MCP server](https://github.com/CycleChain/skillnet-mcp),
  281. maintained by [CycleChain](https://github.com/CycleChain), wraps the SkillNet CLI.
  282. It requires Python, Node.js, and an installed `skillnet-ai` package.
  283. ```bash
  284. git clone https://github.com/CycleChain/skillnet-mcp
  285. cd skillnet-mcp
  286. npm install
  287. ```
  288. Docker:
  289. ```bash
  290. docker pull fmdogancan/skillnet-mcp:latest
  291. ```
  292. Follow the [MCP setup guide](https://github.com/CycleChain/skillnet-mcp#ide--tool-configuration-guide)
  293. to register the server with your agent and check its available tools and
  294. configuration requirements.
  295. ### OpenClaw and JiuwenClaw
  296. SkillNet is available in [OpenClaw](https://github.com/openclaw/openclaw) as a
  297. skill, while [JiuwenClaw](https://github.com/openJiuwen-ai/jiuwenclaw) incorporates
  298. it into its skill marketplace. See the [JiuwenClaw guide](./examples/JiuwenClaw/README.md).
  299. The demo below shows SkillNet running inside OpenClaw to discover and use reusable skills.
  300. <div align="center">
  301. <video src="https://github.com/user-attachments/assets/9d49a00c-827d-47a4-8954-0e6b977ca547" controls="controls" width="800" preload="metadata"></video>
  302. </div>
  303. ---
  304. ## Examples and Experiments
  305. ### Scientific discovery
  306. The scientific workflow notebook illustrates skill discovery and reuse for
  307. scRNA-seq analysis, pathway lookup, and target validation. It uses a predefined
  308. plan and simulated data for parts of the demonstration.
  309. ![Scientific discovery demo](assets/media/scientific-discovery.gif)
  310. [Open the scientific workflow notebook](./examples/scientific_workflow_demo.ipynb).
  311. ### More examples and benchmarks
  312. - [`examples/`](./examples): SDK demos and notebook workflows.
  313. - [`experiments/`](./experiments): reproduction scripts for ALFWorld, WebShop, and ScienceWorld.
  314. Complete the [benchmark environment setup](experiments/README.md) before running
  315. the experiments. Each benchmark has separate environment and data dependencies.
  316. ```bash
  317. cd experiments
  318. python alfworld_run.py --model o4-mini --split dev --max_workers 10 --exp_name alf_test --use_skill
  319. python scienceworld_run.py --model o4-mini --split test --max_workers 5 --exp_name sci_test --use_skill
  320. python webshop_run.py --model o4-mini --max_workers 3 --exp_name web_test --use_skill
  321. ```
  322. ---
  323. ## Roadmap
  324. - Broader evaluation of task routing across local skill libraries.
  325. - More curated skill collections and routing Wikis.
  326. - Stronger skill evaluation and regression testing.
  327. - Extend SkillNet-Fabric routing across skill collections.
  328. - Expand SkillNet-Gym lifecycle evaluation and training environments.
  329. ---
  330. ## Contributing
  331. Contributions are welcome: bug fixes, documentation, examples, integrations, and new skills all help. Please keep pull requests focused and include reproduction steps or examples when possible.
  332. See the [SDK development guide](skillnet-ai/README.md#contributing) for code
  333. organization and local checks.
  334. ---
  335. ## Citation
  336. If SkillNet is useful in your research or agent system, please cite:
  337. ```bibtex
  338. @article{liang2026skillnet,
  339. title={{SkillNet}: Create, Evaluate, and Connect {AI} Skills},
  340. author={Liang, Yuan and Zhong, Ruobin and Xu, Haoming and Jiang, Chen and Zhong, Yi and Fang, Runnan and Gu, Jia-Chen and Deng, Shumin and Yao, Yunzhi and Wang, Mengru and others},
  341. journal={arXiv preprint arXiv:2603.04448},
  342. year={2026}
  343. }
  344. ```
  345. ---
  346. ## License
  347. SkillNet is licensed under [MIT](LICENSE). Skills from external repositories
  348. retain their own licenses.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.