A Claude Code skill that audits a codebase for pieces worth sharing publicly - anywhere from a full standalone library down to a single gist-worthy snippet - and says so honestly when nothing qualifies.
It's built around one filter: a candidate has to be unique and useful, not just cleanly written. Thin wrappers and glue code around a well-known library don't pass. It also checks the current open source landscape for each candidate, so it doesn't call something "unique" that three popular libraries already do better.
Clone this repo into your personal Claude Code skills directory:
git clone https://github.com/<your-username>/open-source-scout.git ~/.claude/skills/open-source-scoutIt'll be available automatically in every project from then on - no per-project setup needed.
Just ask, in whatever project you want audited:
- "What's worth open sourcing in this repo?"
- "Does this codebase have anything worth extracting as a library?"
- "Could any of this be its own open source tool?"
The skill will map the codebase, read the actual implementations (not just file names), check what already exists in the OSS ecosystem for each candidate, and report back three buckets: worth a standalone project, worth a snippet/gist, and not worth it (with reasons - kept as evidence the pass was thorough, not omitted).
MIT - see LICENSE.