history-hygiene

Maintaining a single source of truth in decision tracking requires logging only final outcomes to history.md, not intermediate requests or reversed decisions. This prevents stale or conflicting records from poisoning downstream automated agents.

Logging audit trails for automated agents involves recording final decisions to history.md immediately after confirmation. This approach avoids writing intermediate states, ensuring future spawns read a clean, canonical timeline of actual outcomes.

Stale history causes incorrect decisions because future agents read outdated entries cold, assuming they represent current truth. If a reversed decision is not updated, downstream spawns repeat mistakes based on the wrong version of events.

Updating history.md when a decision is reversed requires immediately replacing the original entry with the final outcome. This ensures the canonical timeline reflects the latest confirmed state rather than leaving conflicting or incorrect intermediate records.

You should not use intermediate states for documentation logging when building workflows for future agents. Writing unconfirmed requests or temporary states creates stale entries that can mislead downstream decision-making processes and break the audit trail.

Yes, you can use history.md for project decision tracking across multiple teams by applying the final-outcome logging approach. It enforces clear guidelines on when to log and how to handle reversals, maintaining a trustworthy shared timeline.