← Back to leaderboard
GitHub API
52
D
GitHub's 1,186-operation spec scores a D, dragged down by two critical failures: operationIds appear machine-generated (often unreadable IDs like `repos/list-for-org`), and 91% of parameters lack descriptions. Pagination scores the lowest at 4% — meaning agents have almost no way to reliably page through the largest API on the planet without burning context on guesswork.
Category breakdown
Examples48
Semantics89
Intent29
Error Handling45
Parameters9
Pagination4
Key findings
- All 1,186 operations have missing or machine-generated operationIds
- 916 of 929 schemas lack examples
- 3,089 of 3,405 parameters have no description
- 265 list operations lack pagination parameters
- 382 operations omit error response documentation
What GitHub should fix
- Replace machine-generated operationIds with human-readable verbs that match the documentation (`listOrganizationRepositories` not `repos/list-for-org`).
- Document the 3,089 undescribed parameters. "limit", "page", "since" — none of these are self-explanatory at the scale of 1,000+ operations.
- Add explicit pagination metadata (next/prev links or cursor parameters) to the 265 list operations missing it.
- Add examples to the 916 schemas without them. With nearly 1,000 schemas, agents need anchors.
How does your API score?
Run AgenticScore on your own OpenAPI spec — find issues before LLMs do.
npx agenticscore score ./openapi.yaml
Get API Key →