| name | deep-research |
| description | Activate comprehensive exploration mode. Use when user says 'deep research' or when a task requires thorough investigation of both internal codebase and external sources before proceeding. |
Deep Research Mode
DEEP RESEARCH MODE — Comprehensive exploration requested.
Required actions:
- Fire at least 3-4 parallel agents before proceeding (they run in background automatically)
- Search BOTH internal (Explorer) AND external (Librarian) sources
- Explore multiple angles: implementations, tests, patterns, docs
- DO NOT proceed with implementation until you have comprehensive context
- Synthesize all findings before making decisions
Research pattern (all Agent calls in ONE message):
// All run in background automatically — fire and continue
Agent(subagent_type: "explorer", description: "Find implementations", prompt: "...")
Agent(subagent_type: "explorer", description: "Find related tests", prompt: "...")
Agent(subagent_type: "librarian", description: "Find official docs", prompt: "...")
Agent(subagent_type: "librarian", description: "Find OSS examples", prompt: "...")
// When all agents complete -> retrieve, synthesize, then act
Now execute deep research for: "$ARGUMENTS"
Fire at least 3-4 agents in parallel covering both internal codebase exploration and external documentation/examples. Synthesize all findings before taking any action.
