| name | journal |
| description | Maintain a compact evidence-backed decision trail for long-running, unattended, or high-risk engineering work. Use when later reviewers must understand what changed, why, and what proved it. Do not activate for routine short tasks or duplicate ordinary progress narration. |
Journal
Record decisions that affect trust in the result, not a transcript of activity.
Decide what deserves a record
Log a choice when alternatives existed, a hypothesis changed the investigation, a risk was accepted, a user constraint redirected work, or evidence changed the plan. Skip routine reads, commands, and obvious mechanical edits.
Also log a material change to an estimate, commitment, scope, dependency owner, escalation threshold, or fallback when later reviewers would otherwise misread what was promised and why it changed.
Record the decision
Each entry must include time or sequence, decision, reason, evidence, affected scope, result, and status. Link stable artifacts or commands instead of pasting large outputs. Mark superseded entries without deleting their history.
When the entry changes a commitment, identify the stakeholder decision, the previous and revised forecast or scope, and the evidence that triggered the change.
Keep it reviewable
Use one row or compact block per decision. Keep facts separate from inference. Do not store secrets, personal data, credentials, or unredacted production content.
Boundaries
Keep the journal local unless the user or repository requires committing or publishing it. Treat that request or trusted project rule as authorization for the ordinary external write and do not ask again. A journal supports verification but does not replace it.
Output
Return the current decision trail or append it to an authorized destination, followed by unresolved decisions and the evidence still needed.
