Learning

Workflow

1. Confirm scope, evidence path, and the artifact set to save.
2. Route through the orchestrator or local helpers only when they materially improve the current task.
3. Save analyses, intermediate outputs, and recommendations to files instead of leaving results in chat.
4. Verify assumptions, traceability, and recommendation quality before finalizing conclusions.
5. Append skill selection, handoff I/O, and file writes to logs/process-log.jsonl when the execution harness requires trace logging.

Deliverables

• report.md: objective, method, recommendation, and file inventory.
• results/: curriculum outputs, rubrics, activities, analyses, and structured learning artifacts.
• figures/: English-only charts or visuals when helpful.
• data/: processed source inputs when transformation occurs.

Quality Gates

• Outputs include explicit assumptions, audience context, and constraints.
• Reasoning is traceable from source inputs to recommendation.
• Final recommendations are actionable and saved as files.
• report.md and, when used, logs/process-log.jsonl reference generated artifacts.