This repository is a collection for publishing and managing Skills for AI agents.
Each Skill is a reusable package of knowledge, workflows, and tools designed to execute specific tasks.
Skills are independent by folder and can be used or installed individually.
The objectives of this project are:
- Publish practical Skills for AI agents
- Design them as reproducible workflows
- Automate tasks such as scripting, analysis, and data fetching
- Experiment with and share the Skill ecosystem
Many Skills are designed as a Python script + explicit processing pipeline.
Example:
fetch ↓ validate ↓ extract ↓ analyze ↓ report
By separating processes in this way, we aim to design Skills that:
- Are easy to debug
- Are highly reusable
- Prevent erroneous inferences by the AI
In this repository, Skills are managed by folder.
A Skill that fetches trending skills from skills.sh and generates:
- Skill rankings
- Keyword analysis
- Developer rankings
- Ecosystem analysis
It can be used for trend analysis and observing the Skill ecosystem.
Folder:
skills/skills-trending-analysis/
stream-crawler is an implementation designed to fetch pages with infinite scroll or lazy loading in stages, prioritizing fetch accuracy over speed.
It accurately captures content on URLs, including static pages, SPAs, and virtual lists, with a reproducible fetching strategy.
Folder:
skills/stream-crawler/
A meta-skill designed to guide AI agents on how to officially publish a newly developed skill to the public repository. It provides experiential best practices, automated structure review, clean copying, and English translation instructions.
Folder:
skills/publish-skill/
Each Skill basically has the following structure:
skills// ├─ SKILL.md ├─ README.md ├─ requirements.txt ├─ scripts/ ├─ references/ ├─ evals/ └─ examples/
Execution specifications for AI agents.
Defines:
- What the Skill does
- How to execute it
- Input/output specifications
- Error handling
Explanation for humans.
Scripts that perform the actual processing.
Files for evaluation and validation of the Skill.
The Skills in this repository are designed with the following principles in mind:
Deterministic processing, such as numerical and statistical processing, is handled by code.
The AI is mainly responsible for:
- Summarization
- Interpretation
- Explanation
This ensures:
- Reproducibility
- Accuracy
- Stability
Skills that handle external data always follow this structure:
fetch ↓ validate ↓ extract
If HTML structures or other data formats change, the Skill halts extraction to prevent the generation of incorrect data.
Skills are designed to run in an independent execution environment as much as possible.
In most cases, it uses:
- Python
- Virtual environment
.venv - requirements.txt
Each Skill can be installed using the npx skills add command.
npx skills add tmiyano89/skillsnpx skills add tmiyano89/skills --skill skills-trending-analysisMIT License
t.miyano
GitHub: https://github.com/tmiyano89