The state of repos
What the corpus says: how repos score on agent readiness, and where they make AI agents stumble.
- Repos judged
- 63
- Average score
- 74
- Median score
- 75
Score distribution
Latest score per repo, bucketed into the bands the judge actually grades by.
- Genuinely excellent (90–100)2
- Solid, with gaps (70–89)52
- Rough (50–69)6
- A real problem (0–49)3
Where repos stumble the most
Category averages across the corpus, weakest first — plus how many prioritized issues landed in each.
| Category | Average score | Issues found |
|---|---|---|
| Reproducible Setup | 64 | 33 |
| Verification Surface | 68 | 28 |
| Context Economy | 69 | 31 |
| Code Navigability | 71 | 33 |
| Agent Safety | 72 | 30 |
| Machine Orientation | 74 | 27 |