research

Multi-subtopic research workflows are coordinated using parallel subagents that explore distinct subquestions, followed by structured synthesis to generate a consolidated final report. A prepared scope.md file initiates the six-phase pipeline to manage the entire investigation.

Provenance tracking preserves the lineage of findings using frontmatter and a derives_from chain throughout the investigation. This structured validation ensures the final synthesized document in the final/ directory remains fully auditable and reproducible.

AI-generated research findings are validated using AVFL-style checks and practitioner Q&A phases during the workflow. These vetting steps require careful configuration of sub-questions and validation parameters to ensure credible, high-confidence results.

Gemini triangulation is natively supported for cross-referencing findings, with CLI fallbacks available when primary validation methods are inaccessible. Both mechanisms operate within the structured synthesis pipeline to maintain evidence notation throughout the final document.

You must prepare a scope.md file defining your research boundaries, sub-questions, and specific AVFL vetting steps before starting the pipeline. Careful configuration of these parameters is required to ensure the parallel subagents produce credible, structured results.

Automated subagent orchestration should be avoided when a topic lacks sufficient scope for parallel exploration or when strict AVFL-style validation cannot be properly configured. Without careful setup of sub-questions and vetting steps, the synthesized report may lack necessary credibility.