PThorpe92/fossier ? reverse-engineered prompt

Build me a GitHub spam prevention tool for open source repos.

I want it to work mainly as a GitHub Action that runs when a pull request is opened or updated, checks whether the author is trusted, known, or suspicious, and then automatically allows, labels for review, or closes the PR. Trusted people should pass through, known contributors should be remembered, and unknown contributors should get scored using signals like account age, repo activity, open PR spam, PR description quality, commit verification, and other obvious spam patterns. If the score is low, it should deny the PR and leave a helpful comment. If it is borderline, it should ask for review instead.

Also make a CLI so maintainers can test decisions locally, vouch or denounce users, scan open PRs, view history, and initialize config files and workflows. Support a local config file and a local database, and include flood detection plus an option to reject AI authored commits. Look up current docs online if you need to.