Implement From Plan

Implement From Plan

Before editing any code:

1. Read AGENTS.md for project context and operating rules.
2. Read ai/plan.md for the scoped task.
3. Internally restate the goal in one sentence before proceeding.

While implementing:

4. Modify ONLY files listed in the "Allowed Files" section of ai/plan.md.
5. Do NOT make unrelated refactors, add unrelated features, or touch files outside scope.
6. Do NOT modify .env, secrets, credentials, or CI configuration unless explicitly allowed.
7. Keep changes minimal — prefer the smallest diff that meets the completion criteria.
8. Follow existing code style (Python: ruff format, 100 char lines, double quotes).

After implementing:

9. Run the verification commands listed in ai/plan.md when feasible.
10. Write ai/codex_report.md with these exact sections:
    • Summary (2-3 sentences)
    • Files Changed (every file touched)
    • Verification Commands Run (commands + exit codes)
    • Results (what worked, what didn't)
    • Risks (potential regressions)
    • Open Questions (anything unclear)
11. Stop. Do not widen scope or start additional tasks.

Source: jam-sudo/Omega — distributed by TomeVault.