What problem does it solve?
Research questions that require multiple independent perspectives are slow, brittle, or produce narrow conclusions when handled sequentially. /research-fleet decomposes a topic into 3–5 independent angles, runs parallel scout agents, and compresses their findings into a single, evidence-backed brief so teams get faster, broader, and more robust answers.
Core Features & Use Cases
- Decomposition: Automatically breaks broad questions into independent, answerable angles or accepts user-provided angles.
- Parallel Execution: Spawns up to five scout agents per wave that run concurrently and write independent findings to a designated planning directory.
- Compression & Iteration: Compresses wave outputs into a unified brief, detects consensus/conflicts/gaps, and optionally launches a targeted Wave 2 to resolve issues.
- Safety & Limits: Enforces scout timeouts (15 minutes), maximum waves, and explicit WebFetch restrictions to avoid unreliable sources.
- Use Case: Evaluate competing technologies, run time-sensitive multi-perspective investigations, or audit a topic that naturally decomposes into independent subquestions.
Quick Start
Use /research-fleet to decompose "Should we migrate from Express to Fastify?" into four scout angles, run parallel scouts, compress their findings, and write the consolidated report to .planning/research/fleet-migrate/REPORT.md.
