eval

Use after a claim-bearing run completes. Writes the docs/evaluation/ write-up, promotes a claim card under docs/claims/ if paper-bound, and links the preregistration + artifact paths.

Use when creating, moving, reviewing, or documenting APXM benchmark, evaluation, session, compiler-diagnostic, evidence, or vLLM-run artifacts. Enforces that generated artifacts live in the repo-local .apxm workspace rather than examples or docs source trees.

Use before any claim-bearing APXM run (benchmark, evaluation, paper-bound number). Drafts, commits, and verifies a docs/preregistrations/ entry under the project's append-only norm.

Use when running the APXM priority-lane benchmark (foreground vs background concurrency). Enforces preregistration, paired-arm cache-salt scoping, and artifact placement under .apxm/.

Use when running the APXM review-council benchmark (multi-judge review). Enforces preregistration, model-zoo deploy order, and artifact placement under .apxm/.