deep-research

Automate educational document creation by guiding AI agents through Scout, Decompose, Research, Assemble, and Verify phases. This workflow generates tiered documents with explicit citations at every level, reducing manual research and synthesis effort.

The multi-agent research workflow is a five-phase process: Scout, Decompose, Research, Assemble, and Verify. It organizes questions and tasks within a planning workspace to produce structured, testable artifacts with strict citation traceability.

Organize research artifacts using a planning directory structure like planning/<question-slug>/. The workflow writes intermediate files such as 00-scout.md and NN-*.md during execution, compiling them into a final FINAL.md document for traceability.

Yes, single-agent workflows are supported alongside multi-agent configurations. Both modes utilize the same planning workspace, artifact lifecycle, and formal verification step to ensure disciplined citation and traceability across the generated educational documents.

No external dependencies or specific tool components are required to run this research workflow. It operates independently using a file-based artifact lifecycle and a structured planning workspace to manage the generation and verification process.

Use a tiered document generation workflow when you need progressive understanding with explicit citations at every tier and formal verification. It is ideal for educational materials requiring traceable artifacts, unlike standard AI research that may lack structured verification.