← All modules Repo Radar
Scraped
505
Build on
4
Trial
18
Untriaged
400
Skipped
77
Read me How we actually use other people's repos

1 · It arrives as a skill or an MCP server

The good ones aren't apps we open — they install into this setup and Claude calls them. npx skills add …, pip install … plus an MCP entry, or a symlink into ~/.claude/skills/. After that you just ask for the edit in plain English, in the same chat you brief from.

2 · Trial it on our own footage, in 45 minutes

Never the repo's demo file — a real swipe-gallery clip or a real bottle shot. Everything looks good on its own demo. If it hasn't produced something by the 45-minute bell, that is the result: it costs too much to set up.

3 · The licence decides what we're allowed to do

MIT / Apache — use it, change it, ship it. GPL / AGPL — fine for a tool we run here, a problem inside anything customer-facing. No licence file means all rights reserved: we can read it and copy the idea, never the code.

4 · Pin it, then depend on it

A repo that ships daily will break a machine on its own schedule. Whatever wins gets locked to one version, and the machine's README says which. That's the difference between using someone's repo and inheriting their bugs.

⚠ What "Trust" here is and isn't

It's a provenance read — licence, ownership, age, contributor count, release discipline, whether it's still alive. It is not a security audit. Nobody has read the source. Treat anything below Reasonable as code you wouldn't run outside a throwaway folder.

Where the list comes from

A sweep of 505 repos and counting, across 21 search axes tied to the machines we already run — plus anything scraped off X. Stars are shown but score nothing: they measure attention, not fit.