investigation-workflow

System investigation uses a structured six-phase workflow to explore architecture, dependencies, and behavior. The phases scope the analysis, plan exploration strategy, deploy parallel deep-dives into components, synthesize findings into documentation, validate outputs, and transfer knowledge—enabling efficient understanding of complex systems without scattered notes.

Yes. The workflow deploys parallel agent-driven deep-dives, allowing you to investigate authentication flows, CI failures, and architecture patterns at the same time. Each dive operates independently within a bounded token budget, then results merge into unified documentation and validation.

The investigation workflow captures findings through automatic knowledge synthesis and documentation in Phase 4, then validates accuracy in Phase 5 before transferring to stakeholders in Phase 6. This ensures documentation stays synchronized with actual system behavior rather than relying on manual notes.

Phase 1 explicitly defines scope and boundaries; each parallel dive operates within a separate, bounded token budget. Explicit triggers and phase-based deliverables prevent scope creep while distributed exploration maximizes efficiency across multiple concurrent investigations.

Structured investigation works best when you need repeatable outputs, multiple team members exploring simultaneously, explicit documentation of findings, or analysis of complex systems like authentication flows or CI pipelines. Ad-hoc exploration suits quick questions; workflows suit systematic understanding.