Source provider policy

GitHub API source discovery

This page documents the operator-facing source pipeline. Public loop pages do not automatically republish raw files. Candidate files are fetched, scored, optionally classified, and kept in an admin review queue before any loop can become public.

Discovery

Search queries run in controlled batches through the server-side source discovery pipeline.

Candidate storage

Candidate metadata, excerpts, hashes, classifier output, and review status are stored in D1-compatible tables.

Human review

Candidates require review before publishing. Raw source material is not treated as original site content.

Publishing rules

Discovery results enter the admin queue as candidates, not public pages.
AI classification is advisory. It cannot publish a loop by itself.
Published loops must add structure: goal, exit condition, steps, guardrails, expected output, attribution, and safety notes.
Low-value copies, unsafe commands, unclear licenses, and generic prompts should be rejected.

For the public user-facing explanation, see how source discovery works.