Evals And Claims
The public CLI repo includes deterministic product tests and small synthetic fixtures. It is the product claim surface, not a dump of exploratory research material or unreduced evaluation runs.
Public claims should stay tied to:
- behavior visible in the CLI;
- reproducible public fixtures;
- demo transcripts generated from current commands;
- documented command output;
- privacy and storage guarantees that are true in the public repo.
For the current CLI boundary, see the public repository's EVALS.md.