What problem does it solve?
Prevents data loss, context truncation, and AI hallucination by enforcing a strict file-saving workflow for analysis outputs so downstream agents and coordinators can reproduce and validate results.
Core Features & Use Cases
- Mandatory persistence: Requires every analysis output to be written with the Write tool to a coordinator-provided output_path rather than returned only in prompts.
- Dual-format storage: Enforces saving both machine-readable JSON and a human-readable Markdown summary with numeric filename prefixes for ordering and traceability.
- Robust error handling & session resume: Defines explicit SUCCESS/FAIL responses with error codes (PATH_NOT_FOUND, PERMISSION_DENIED, DISK_FULL, UNKNOWN_ERROR) and includes session resume verification that checks for required JSON files before progressing.
- Use Case: Orchestrating a DC pension portfolio analysis where multiple agents (index-fetcher, rate-analyst, sector-analyst, etc.) each produce JSON outputs and coordinated numbered markdown summaries consumed by downstream agents.
Quick Start
Use the Write tool to save the analysis_result as portfolios/{session_folder}/rate-analysis.json and also write a numbered markdown summary 00-rate-analysis.md into the same output_path.