.omo/evidence/20260721-ulw-loop-gajae-adoption/README.md
This directory records evidence for the approved ulw-loop gajae-adoption plan:
steer --proposals-json,task-1.md through task-5.md.80a57345f: rejected batches append exactly one indexed steering_rejected audit, accepted batches use one plural ledger append for fresh audits, conflicting CLI flags fail with ULW_LOOP_STEERING_BATCH_CONFLICT, and validation-batch branch failures use the approved distinct typed codes.~/.codex/config.toml SHA-256 was unchanged by the built-CLI QA and reviewer-fix QA.task-6.md, task-6-e2e-transcript.txt, and reviewer-fix-transcript.txt.reviewer-fix-report.md and reviewer-fix-transcript.txt supersede earlier claims where blocker details changed.The artifacts cover the unit/component gates, the repo Codex gate, real local-build CLI behavior for all three adopted features, fail-closed validation-batch behavior, independent-review blocker fixes, isolation/cleanup requirements, and restoration of an out-of-scope generated CodeGraph diff after the root gate.
No PR was opened/pushed/merged. No published package was used. No real Codex plugin cache was used. Live app-server/TUI Codex QA was omitted because the approved todo-6 QA surface is the component's built CLI in an isolated temp git repo. The accidental .omo runtime files are removed in the final evidence commit; only component changes and .omo/evidence/** remain intentional deliverables.
task-1.mdtask-2.mdtask-3.mdtask-4.mdtask-5.mdtask-6.mdtask-6-component-gate.txttask-6-codex-gate.txttask-6-e2e-transcript.txtuncommitted-diff-salvage.mdreviewer-fix-report.mdreviewer-fix-transcript.txtf1.mdf2.mdf3.mdf4.md