rafaie/codex_home/debug-loop

1) Reproduce the failure (exact command, inputs, environment). 2) Minimize to the smallest failing case. 3) Add a regression test that fails before the fix. 4) Fix the root cause. 5) Re-run the exact reproducer from step 1 and require it to pass. 6) Run the narrow regression test, then the appropriate command from codex.toml:

• implementation defect: test_quick, else test, else uv run pytest -q
• final/full-gate defect: test_full, else test, else uv run pytest -q
• smoke/runtime/packaging defect: rerun smoke-test and record its new artifact path

7) Update the work item docs with repro, root cause, fix, regression test, and verification evidence.

• Prefer test-results.md for failures/fixes.
• Use implementation.md for implementation notes.
• For legacy single-file specs, update the Debug log section.