skills/orchestrate-coding-agent-execution/assets/checkpoint-report-template.md

# Execution checkpoint

Project / checkpoint ID / recorded time:
Task / attempt / phase / gate / execution state:
Agent / provider / relevant model-mode / session identifier if safe and available:
Current repository / branch / revision / dirty tree and ownership:
Active source manifest / harness version / applicable instructions:
Changed files:
Approved Class-2 decisions, rationale and evidence:
Tests actually run, results, revision and evidence IDs:
Assurance mode and rationale:
Acceptance oracle(s), direct observation and non-circularity check:
Acceptance environment and identity, or N/A with reason:
Visual status and last approved visual checkpoint:
Independent review / owner approval records:
Known failures and unverified claims (separate from confirmed facts):
New risks / open Class-1 decisions / blockers:
Last approved checkpoint / restore scope and authority:
Context health / session strategy / rationale:
Next safe task and dependency source:
Next task contract location/version:
Remaining approvals and exact stop condition:

Update only an authorized state artifact. Preserve failed attempts and history; do not overwrite confirmed state with an unverified agent report. No passwords, API keys, tokens or sensitive session credentials.