| name | prior-art-search |
| description | Use when asked to find similar projects, existing tools, prior art, libraries, or whether something already exists before building. |
| metadata | {"short-description":"Multi-angle search for existing tools before building"} |
Prior-Art Search
One search angle always misses things. Sweep several, then judge.
1) Sweep (parallel angles)
- By problem words: GitHub search on 2–3 phrasings of the problem, not the imagined solution name.
- By ecosystem:
npm search / PyPI / crates for the domain keywords; registries surface things GitHub search ranks poorly.
- By neighbors: who depends on / is depended on by the closest known tool (
gh api dependents, awesome-lists in the domain).
- By feature: search issues/discussions of the big incumbent (e.g. Playwright, ESLint) for people asking for this feature — links to third-party solutions cluster there.
2) Qualify each candidate
Per candidate, one row: stars, last commit date, open-issue responsiveness, license, and — most important — the exact overlap and the exact gap vs the idea. Install/run the top 2 candidates on a real example when feasible; READMEs oversell.
3) Conclude
- Adopt (which one, and the gap you accept), extend (fork/plugin point), or build (state what specifically nothing covers).
- List dead/abandoned lookalikes too, with why they died — that's design input, not noise.
Notes:
- Timebox: if 4 angles surface nothing close, say so and stop — absence of prior art after a real sweep is a valid, reportable result.
