Finding (surfaced in malaria-rdt/20260606-0719 Phase 4)
The connect-program-setup-eval active_window_status dimension still anchors on "opp status must be draft at end of Phase 4." But connect-opp-setup Step 6.5 now activates the opportunity synchronously by design (so Phase 6 app-screenshot-capture has a real signed-in user against an active opp; opp is created is_test=true so prod analytics exclude it).
The eval rubric's anchor predates the activation move, so it can fail a correctly-activated opp. The eval graded around it (scored 8.0 on that dim, no hard-gate) rather than fail on the stale anchor.
Fix lands here
skills/connect-program-setup-eval/SKILL.md (or wherever the active_window_status rubric anchor lives) — refresh the anchor to expect active (with is_test=true) at end of Phase 4, matching connect-opp-setup Step 6.5's synchronous activation.
Repro
malaria-rdt run 20260606-0719 — connect-program-setup-eval verdict, active_window_status dimension.
Finding (surfaced in malaria-rdt/20260606-0719 Phase 4)
The
connect-program-setup-evalactive_window_statusdimension still anchors on "opp status must bedraftat end of Phase 4." Butconnect-opp-setupStep 6.5 now activates the opportunity synchronously by design (so Phase 6 app-screenshot-capture has a real signed-in user against an active opp; opp is createdis_test=trueso prod analytics exclude it).The eval rubric's anchor predates the activation move, so it can fail a correctly-activated opp. The eval graded around it (scored 8.0 on that dim, no hard-gate) rather than fail on the stale anchor.
Fix lands here
skills/connect-program-setup-eval/SKILL.md(or wherever theactive_window_statusrubric anchor lives) — refresh the anchor to expectactive(withis_test=true) at end of Phase 4, matching connect-opp-setup Step 6.5's synchronous activation.Repro
malaria-rdt run 20260606-0719 — connect-program-setup-eval verdict, active_window_status dimension.