source-package-internals
Skill1 starsChanged 23 days agoMITView on GitHub
How to use a technology's own repository at the pinned version as a codelabgen source — examples, tests, docstrings, notebooks, cookbooks and templates — plus LLM-oriented docs (llms.txt, Context7). The strongest code-truth material. Source type `package-internals`.
In this repository24 other files it gives its agents.
- code-authoringSkill
- codelabgen-orchestratorSkill
- course-extensionSkill
- curriculum-designSkill
- diagrammingSkill
- example-designSkill
- lesson-reviewSkill
- notes-writingSkill
- readalong-writingSkill
- runtime-strategySkill
- site-assemblySkill
- source-articlesSkill
- source-books-coursesSkill
- source-digestSkill
- source-evaluationSkill
- source-github-reposSkill
- source-official-blogSkill
- source-official-docsSkill
- source-papers-specsSkill
- source-qaSkill
- source-talks-workshopsSkill
- source-universitySkill
- state-protocolSkill
- topic-extractionSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
Check version_matched in its output.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
The library itself (package-internals)
Nothing shows how an API really behaves in the pinned version better than the library's own code. Tests are executable specifications, examples are maintained by the authors, and docstrings explain intent. This is the course's best code-truth.
Get the code at the pinned version
- Find the repo from
research/overview.mdorcodelabgen pkg-info. codelabgen repo clone <owner/name> --package <package name>. It finds the tag of the pinned version (including monorepo tags likelanggraph==1.2.15) and prints the interesting folders.- Check
version_matchedin its output. If it is false, the clone is the default branch: say so in eachreason, and lower the score a little.
What to look at
| Where | Why |
examples/, cookbook/, templates/, notebooks | Complete, maintained usage examples |
tests/ of the main package | How each API behaves, including edge cases and errors. Find a feature's tests by grepping for its name |
| Public modules' docstrings | Intended usage and parameter meanings |
docs/ in the repo | Docs sources, sometimes with more detail than the site |
| Official template or starter repos (often separate repos in the same org) | Project layout as the maintainers recommend it |
Also check for LLM-oriented docs: codelabgen llms-txt <docs root>. If the Context7 MCP tools are available in your session, look up the library there too.
What to record
Record GitHub URLs pointing into the repo at the pinned tag (https://github.com/<repo>/tree/<tag>/<path>), one per useful folder or file group:
- the examples or cookbook root (
code-truth, oftenstructure) - the tests of the core package (
code-truth, andpitfallswhen they cover error cases) - notable official templates (
code-truth) llms.txt, if it exists and the official-docs hunter hasn't recorded it (structure,code-truth)
In reason, say what is there and how to use it: "Tests for StateGraph compile errors: unknown nodes, missing entry point, duplicate names."
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

