Files
Agent-Intent/evaluation
agent-exec-transport 7831228114 Document the baseline-drift caveat and harden the slice assertion
Why: A3 review minors - BASELINE.md now states engine changes do not
trigger baseline invalidation alarms (manual re-run required), and the
missing-slots assertion uses explicit element-wise comparison.
Evidence: eval re-run matches committed metrics (verify-min-int2.json,
exit=0); corpus digest unchanged.
Impact: doc caveat + assertion helper; no behavior change.
Refs: feature/audit-review-minors-cleanup
2026-10-23 10:06:07 +00:00
..