| name | research |
| description | Deep research on a topic using web search and analysis capabilities |
| tags | ["research","search","analysis","web"] |
| version | 2.0.0 |
Research
Conduct deep research on a topic. Research agent owns all web search for the fleet.
When to use
- "Research this topic"
- "What are the current best practices for X?"
- "Find out about this company/tool/service"
- "Fact-check this claim"
- Any agent needing web search routes through Research via Captain
Process
- Search Chartroom first — we may already know this
- Formulate a clear research question
- RUN research-estimate FIRST — mandatory cost gate
- If estimate is over $0.50: STOP, send approval choices to human via Captain/Relay
- For web search: use the
web-search skill (queues Gemini CLI on host, free tier primary, paid Flash Lite failover)
- Verify key claims across multiple sources
- Structure findings: what we learned, confidence level, sources, impact on us
- Report results back to the requesting agent and/or user via Captain/Relay
Status Reporting
- On start: "Researching [topic] via [Gemini search/multimodal/etc]"
- On complete: Structured findings + sources + confidence + token cost
- On failure: What broke + token cost of failed attempt
Cost Discipline
- research-estimate is MANDATORY before expensive operations
- Web search via
web-search skill: free (Gemini CLI free tier) or pennies (paid Flash Lite failover)
- Under $0.05: execute, report cost
- $0.05-$0.50: execute, flag cost prominently
- Over $0.50: DO NOT execute — send choices to human via Relay buttons
- Track spend in workspace MEMORY.md
Output Format
## Research: [Topic]
**Question**: [What was asked]
**Findings**: [Key points, numbered]
**Sources**: [URLs with trust ratings]
**Confidence**: [High/Medium/Low + why]
**Impact on us**: [What this means for OpenClaw]
**Suggested action**: [If any]
**Token cost**: [Estimated spend for this research]
Intent: Informed [I18]. Purpose: [P-TBD].