STORM Research

This skill turns a loose topic into a research structure. It prepares writing; it does not write the article.

Scope

Do:

• scan multiple perspectives
• identify conflicts, consensus, and blind spots
• write a research brief
• score confidence
• create an evidence plan for wenchang-research
• update content_state and handoff

Do not:

• write a final article
• treat simulated perspectives as facts
• replace source-backed fact checking
• make medical, legal, financial, or high-stakes decisions
• use unsupported social screenshots as primary evidence

Process

1. Run five default perspectives:
   • practitioner
   • researcher
   • skeptic
   • economic observer
   • historical observer
2. Build a contradiction map:
   • direct conflicts
   • shared assumptions
   • strongest evidence
   • weakest evidence
   • blind spots
3. Write a research brief:
   • one-paragraph summary
   • 5 key findings
   • hidden connection
   • actionable insight
   • frontier question
4. Review confidence:
   • score each finding
   • identify weakest claim
   • list missing perspectives
   • list facts requiring verification
5. Produce an evidence plan for wenchang-research.

Output Format

## Research Decision
- Fit for next stage:
- Recommended next step:
- Main reason:

## Perspectives
| Perspective | Core concern | Supports | Challenges | Unique information | Needs verification |
| --- | --- | --- | --- | --- | --- |

## Contradiction Map
- Conflicts:
- Consensus:
- Strongest evidence:
- Weakest evidence:
- Blind spots:

## Research Brief
- Summary:
- Key findings:
- Hidden connection:
- Actionable insight:
- Frontier question:

## Confidence Review
- Scores:
- Weakest claim:
- Bias check:
- Missing perspectives:
- Must verify:

## Evidence Plan
1. <question>: <suggested primary sources>

## content_state Update
```yaml
content_state:
  storm_research:
    topic:
    perspectives: []
    contradiction_map:
      conflicts: []
      consensus: []
      blind_spots: []
    confidence_review:
      verification_needed: []
    evidence_plan: []
  next_step:
    skill: wenchang-research
    reason:
    user_decision_needed:
  handoff:
    from_stage: storm-research
    to_stage:
    accepted_inputs: []
    ignored_context: []
    stop_condition:

## Stop Conditions

Pause if there are more than three equally strong angles, if the topic involves high-stakes advice, or if core claims cannot be verified with primary sources.

## 作者入口

- 微信公众号：`AI生命克劳德`
- X：[@yangchao228](https://x.com/yangchao228)
- GitHub：[yangchao228](https://github.com/yangchao228)