persisting-agent-outputs

What problem does it solve?

Agent outputs must survive session resets and remain discoverable by the orchestrator to enable task resumption, auditing, and review.

Core Features & Use Cases

• Output Location Protocol: All agent outputs go to a designated path such as .jonggrang/.output/features/{feature_id}/{phase}-{role}-output.json.
• Standardized header: Each output file includes a header with keys jonggrang-output, feature_id, phase, role, task_id, timestamp, status, and a payload section for role-specific data.
• Role-specific payloads: Lead (architecture plan), Developer (implementation), Reviewer (review report), and Tester (test results) produce structured payloads under output.
• Use Case: After completing a task, an agent writes a verifiable log that allows the orchestrator to resume work after restarts and to audit the decision path.

Quick Start

Store the first architecture plan output to the designated path using the standard header.