oss-skill

For Chinese, see README-zh.md.

Turn engineering judgment from open-source projects into reusable skills.

oss-skill studies how strong open-source authors, repositories, and maintainer teams make engineering decisions, then turns those patterns into reusable skills for coding agents like Claude Code and Codex.

What it does

• Reads code, commits, PRs, issues, release notes, and design docs
• Extracts how a project or maintainer makes decisions
• Generates reusable skills for coding, review, refactoring, API design, debugging, and maintenance decisions

The output usually looks like this:

skills/<target>-playbook/
├─ SKILL.md
├─ scripts/
└─ references/
   └─ research/

Install

npx skills add lianchi/oss-skill

Then use it in your agent:

> Distill the Vite project
> Distill the vercel-labs team
> Review this patch by React maintainer standards

After a run, skills/<target>-playbook will be generated in your current project. Copy that folder into your agent's skills directory, for example ~/.codex/skills/.

When to use it

Good fits:

• Distilling an author, repository, or maintainer team
• Reviewing a patch with a project-specific standard
• Learning engineering habits from mature open-source projects
• Turning maintainer judgment into a reusable team skill

Not a good fit:

• Personality imitation
• Non-software topics
• Requests with no clear engineering goal

How it works

1. Pick a target, or start from a problem and work backward to the right target.
2. Research it with a code-first approach.
3. Organize the evidence across 6 dimensions.
4. Turn the result into an executable skill.

The 6 research dimensions are:

1. Architecture
2. Commits and PRs
3. Issues and reviews
4. Docs, talks, and public writing
5. Key decision cases
6. Evolution and lineage

The final skill should help with real development work, such as:

• feature implementation
• bug fixing
• patch review
• refactoring
• API design
• performance optimization

Why this exists

AI is getting very good at writing code, and general coding agents are already good at producing complete solutions.

What oss-skill tries to strengthen is another layer: turning the engineering judgment repeatedly proven in strong open-source authors and mature repositories into skills that can still be reused today.

Inspired by nuwa-skill

oss-skill is clearly inspired by alchaincyf/nuwa-skill in its overall approach, structural organization, and implementation, and it also directly reuses some of its practices, such as parts of the scripting approach.

It deserves huge thanks and respect. nuwa-skill turned “distillation” into a practical workflow, and helped many people realize that public materials, engineering traces, and structured prompts can be organized into reusable capabilities.

The focus is different:

• nuwa-skill is more about human thinking frameworks and cognitive style
• oss-skill is more about software engineering judgment, code review, architecture tradeoffs, and maintainer decisions