Evidence Tracker Skill

Purpose

Maintain comprehensive log of all research sources, citations, and evidence used in content creation to ensure credibility and traceability.

When to Use This Skill

Invoke when:

• Research skills have gathered data
• Content generation uses facts, stats, or quotes
• User asks to "cite sources" or "track evidence"
• Quality assurance requires source verification
• Throughout research and content workflows

Instructions

Step 1: Initialize Evidence Log

Create evidence tracking structure:

# Evidence Log

**Session:** {workflow_name}
**Topic:** {topic}
**Started:** {date_time}

---

## Sources by Type

### Web Sources (Exa)

- Total: {count}
- Depth: {comprehensive|standard|quick}

### Social Media Research (social-media-mcp)

- Total: {count}
- Platforms: {platforms}

### YouTube Videos (youtube-transcript)

- Total: {count}
- Total duration: {minutes}

### Profiles Analyzed (Apify)

- Total: {count}
- Platforms: {platforms}
- Cost: ${total}

---

## Evidence Items

{Items below}

---

## Confidence Assessment

**Overall Confidence:** {score}/10
**Based on:** {sample_size} sources, {recency} data, {authority} sources

Step 2: Track Each Evidence Item

For each fact, statistic, quote, or claim:

Evidence Entry Format:

### Evidence #{id}: {short_description}

**Claim/Finding:**
"{exact_text_or_data}"

**Source:**

- Type: {web|social|video|profile|news}
- Title: {source_title}
- URL: {url}
- Author/Channel: {if_available}
- Date: {publication_date}
- Timestamp: {if_video} (e.g., 2:34)

**Confidence:**

- Score: {high|medium|low}
- Based on:
  - Authority: {credible_source?}
  - Recency: {how_recent?}
  - Corroboration: {confirmed_elsewhere?}

**Used in:**

- Workflow: {workflow_name}
- Output: {content_item}
- Section: {where_used}

**Retrieved:** {date_time}

---

Step 3: Categorize Evidence

Organize by type:

A. Data & Statistics

• Numerical claims
• Percentages
• Growth rates
• Market size
• Survey results

B. Expert Quotes

• Direct attributions
• Expert opinions
• Industry insights
• Contrarian views

C. YouTube Evidence

• Video quotes with timestamps
• Hook examples
• Structure patterns
• Teaching approaches

D. Profile Evidence

• Engagement metrics
• Top performing posts
• Content patterns
• Platform-specific data

E. Trend Evidence

• Trending topics
• Hashtag volumes
• News articles
• Social momentum

Step 4: Assess Confidence

For each evidence item, evaluate:

Authority (40%):

• Tier 1: Major publications, academic, government (High)
• Tier 2: Industry blogs, verified creators (Medium)
• Tier 3: Unknown sources, unverified (Low)

Recency (30%):

• Last 30 days: High
• Last 3 months: Medium
• Last 6 months: Medium-Low
• Over 6 months: Low

Corroboration (30%):

• 3+ sources: High
• 2 sources: Medium
• 1 source: Low

Final Confidence Score:

• High: 8-10/10
• Medium: 5-7/10
• Low: 1-4/10

Step 5: Link Evidence to Content

When evidence is used in content:

Update evidence entry:

• Add "Used in:" section
• Note which content item
• Track which specific claim

In generated content:

• Add footnote or inline citation
• Format: Source: {title}
• For videos: Source: {title} at {timestamp}

Step 6: Create Evidence Summary

At end of workflow:

## Evidence Summary

**Total Evidence Items:** {count}

**By Type:**

- Data & Statistics: {count}
- Expert Quotes: {count}
- YouTube Evidence: {count}
- Profile Evidence: {count}
- Trend Evidence: {count}

**Confidence Distribution:**

- High (8-10): {count} items
- Medium (5-7): {count} items
- Low (1-4): {count} items

**Source Breakdown:**

- Exa web search: {count}
- social-media-mcp: {count}
- YouTube: {count} videos
- Apify profiles: {count}

**Recency:**

- Last 30 days: {count}
- Last 3 months: {count}
- Older: {count}

**Overall Quality Score:** {score}/10

Step 7: Save Evidence Log

Save to: jarvis-sidecar/sessions/{workflow}-{topic}-evidence-{date}.md

Also update: Research brief or content output with evidence links

Example

Evidence Entry:

### Evidence #1: AI automation market growth

**Claim/Finding:**
"AI automation market growing 28% YoY"

**Source:**

- Type: web
- Title: "State of AI Automation 2025"
- URL: https://techcrunch.com/ai-automation-2025
- Author: TechCrunch Research
- Date: 2025-10-15

**Confidence:**

- Score: High (9/10)
- Based on:
  - Authority: Tier 1 publication (TechCrunch)
  - Recency: 12 days old
  - Corroboration: Confirmed in 2 other sources

**Used in:**

- Workflow: research-topic
- Output: Research Brief "AI Tools"
- Section: Data & Statistics

**Retrieved:** 2025-10-26 14:30:00

---

Quality Standards

• Every data point must have source
• All quotes must have attribution
• Video evidence must include timestamps
• Confidence scores must be justified
• URLs must be captured (not just titles)
• Publication dates required
• Used-in tracking mandatory

Integration Notes

Works with:

• social-media-research (tracks facts, news, trends)
• youtube-research (tracks quotes with timestamps)
• deep-web-research (tracks insights, data, quotes)
• profile-analysis (tracks engagement metrics)
• research-synthesizer (consolidates all evidence)
• Content generation workflows (citations)

Provides:

• Credibility for generated content
• Source verification
• Traceability for claims
• Quality assurance

Edge Cases

Source URL not available:

• Use source title + platform
• Note: "URL unavailable"
• Include retrieval date

Conflicting evidence:

• Track both items
• Note conflict in confidence
• Research further or note limitation

Evidence not used in final content:

• Keep in log (might be useful later)
• Mark as "Not used" in tracking
• Maintain for reference