Published before anyone was scored
Every score point has an anchor. That is what makes exact agreement between validators reachable, and a score against a standard nobody can read is worth nothing.
What agreement means
Two validators mark the same source independently. They count as agreeing when no criterion differs by more than 1, at most 1 criterion differs at all, and the band is identical. The total is summed by the contract, in deterministic code, from the leader's marks. The reasons are not compared - two validators reading the same code write the same judgment in different words.
Counted, or judged
4 of these criteria are counted from the source in deterministic code - agreement - untrusted - boundary - failure - so every validator derives the same score and the same reason without asking a model. The rest are judged: necessity - finality - mechanism - provenance - overreach - recourse. Every mark below says which it is.
No report is issued and the splitting criterion is named. That is a signal the anchor is written badly, our problem rather than the submitter's.
| Criterion | Subject | Splits | Reads as |
|---|---|---|---|
| agreement | contract | - | counted, never judged |
| boundary | contract | - | counted, never judged |
| failure | contract | - | counted, never judged |
| finality | site | 0 | clear |
| mechanism | site | 0 | clear |
| necessity | contract | 0 | clear |
| overreach | site | 0 | clear |
| provenance | site | 0 | clear |
| recourse | site | 0 | clear |
| untrusted | contract | - | counted, never judged |
0 reports - 0 contested - 0 splits recorded - rubric v1