nci-manipulation-analysis

# NCI Manipulation Analysis

This skill uses pattern-based manipulation detection that identifies **how** content tries to influence the reader, not whether claims are factually true. Manipulation techniques leave fingerprints regardless of underlying accuracy.

Use TodoWrite to track these mandatory steps:

<required>
1. Input Processing (text or URL)
2. Score all 20 categories (1-5 scale each)
3. Calculate 5 composite factors
4. Calculate overall score (0-100)
5. Check deep research triggers
6. Generate perspectives (manipulative + legitimate)
7. Output report
</required>

## Quick Start

### For Text Content
1. Read the content provided by user
2. Apply 20-category analysis (see [references/categories.md](references/categories.md))
3. Calculate composite factors and overall score (see [references/scoring.md](references/scoring.md))
4. **Check deep research triggers** - if score > 40 or key categories elevated, verify claims
5. Generate dual perspectives
6. Output report in requested format

### For URLs
1. Use `WebFetch` to retrieve content from URL
2. Extract main article/post text
3. Proceed with text analysis workflow
4. Note source metadata (publication, date, author)
5. **If triggers met**: Use `fact-checker` agent to verify key claims

## First Principles (Summary)

The NCI Protocol is grounded in these principles (see `agents/perspective-generator.md` for full version):

1. **Evidence over authority** - Evaluate patterns in content, not source reputation
2. **Steel-man interpretation** - Present strongest version of each perspective
3. **Atomic decomposition** - Break claims into smallest verifiable units
4. **Source agnosticism** - Apply identical standards regardless of source alignment
5. **Bidirectional beneficiary analysis** - Ask who benefits if believed AND if dismissed
6. **Pattern vs. Intent** - Focus on techniques; deep research evidence can inform motives

These principles ensure fair, consistent analysis across all content regardless of politic