Create detailed research plans by decomposing structured prompts into subtopics, search strategies, and agent deployment configurations. Extracted from Phase 2 of research-executor for standalone planning capabilities.
Takes a structured research prompt (from question-refiner) and creates a comprehensive execution plan with subtopic decomposition, search strategies, and multi-agent deployment configuration.
question-refiner (structured prompt)
↓
research-planner (this skill)
↓
research-executor (validates & executes)
Required: Structured prompt with TASK, CONTEXT, SPECIFIC_QUESTIONS, KEYWORDS, CONSTRAINTS, OUTPUT_FORMAT
Optional: Complexity level, budget constraints, preferred agent types
# Research Plan: [Topic]
## 1. Executive Summary
- Topic, Research Type, Complexity
- Estimated Duration: [15-90 min]
- Estimated Cost: [$X]
## 2. Subtopic Decomposition
[3-7 subtopics with priority]
## 3. Search Strategies
[3-5 queries per subtopic]
## 4. Data Sources
| Source Type | Priority | Rationale |
## 5. Agent Deployment
- Total Agents: [3-8]
- Model Mix: [sonnet + haiku]
- Assignments per agent
## 6. Resource Estimation
| Resource | Estimate |
|----------|----------|
| Time | X min |
| Tokens | X |
| Agents | X |
## 7. Quality Gates
- Phase 3: ≥80% agent success
- Phase 5: ≥30 citations
- Final: Quality ≥8.0
## 8. Approval Options
✅ Approve | 🔧 Modify | 🔄 Alternative | ❌ Cancel
| Research Type | Agents | Model Mix |
|---|---|---|
| Quick Query | 2-3 | All haiku |
| Standard | 4-5 | 2 sonnet + 3 haiku |
| Deep Research | 6-8 | 3-4 sonnet + rest haiku |
| Technical | 3-5 | All sonnet |
Time (min) = 15 + (subtopics × 5) + (agents × 3)
Tokens = agents × 15,000 + 10,000 (overhead)
Users can request:
Upstream: question-refiner (structured prompt)
Downstream: research-executor (execution plan)
Parallel: ontology-scout (domain reconnaissance)
See also: Skill Base Template
See examples.md for planning scenarios.
See instructions.md for implementation guide.