PageProof

Give it an API endpoint and how it paginates (cursor, offset, or page), and PageProof proves -- with evidence, not a guess -- whether it's honest: no duplicates across pages, no stuck or backwards cursor, claimed total matches what was actually retrieved.

Format: the brief must be JSON, not free text -- e.g. {"base_url": "https://api.example.com/items"}. Only base_url is required; items_field, id_field, total_field etc. are optional with sane defaults. Unparseable JSON gets formatting instructions back, not a guess.

The rule that makes this different: no invariant to check completeness against? Returns INCONCLUSIVE, never a false PASS. A clean traversal isn't proof nothing was missed.

Output: PASS / FAIL / INCONCLUSIVE / TIMEOUT / UNREACHABLE + evidence (pages fetched, unique records, duplicate examples, totals at start and end of the run).

Good for: vetting a new integration before building a data sync on top of it, catching a silent break after a vendor API update, a fast check on an undocumented API, a second opinion on pagination code you just wrote.

Tested against a real API before listing: 34-page live traversal, 1,351 records, verified PASS.