Dedup Validator Agent

Validate and deduplicate Phase 1 results.

Purpose

• Remove suggestions that already exist
• Remove duplicates between agents
• Flag conflicts with existing rules
• Prioritize and sort final results

Execution

This agent runs after Phase 1 completes. It receives:

• Results from all Phase 1 agents
• Current project documentation

Validation Rules

1. Exact Match

Content already exists verbatim in documentation. → Status: duplicate, remove from suggestions

2. Semantic Match

Content means the same thing, differently worded. → Status: duplicate, remove from suggestions

3. Conflict Detection

New suggestion contradicts existing rule. → Status: conflict, flag for user review

4. Quality Filter

Suggestion is too generic or obvious. → Status: filtered, remove from suggestions

5. Cross-Agent Duplicate

Same suggestion from multiple agents. → Keep highest confidence version

Prompt

Validate and deduplicate the Phase 1 analysis results.

## Phase 1 Results

### Doc Analyzer
{{docAnalyzerResults}}

### Automation Finder
{{automationFinderResults}}

### Learning Extractor
{{learningExtractorResults}}

### Followup Planner
{{followupPlannerResults}}

## Existing Documentation
{{existingDocs}}

## Validation Tasks

1. **Check for Exact Duplicates**
   - Is this content already in the documentation?
   - Mark as duplicate if exists

2. **Check for Semantic Duplicates**
   - Does existing doc cover the same concept?
   - Mark as duplicate if meaning is same

3. **Check for Conflicts**
   - Does this contradict existing rules?
   - Mark as conflict and note the conflict

4. **Cross-Agent Deduplication**
   - Did multiple agents suggest the same thing?
   - Keep only the best version

5. **Quality Filter**
   - Is the suggestion specific and valuable?
   - Remove generic or obvious suggestions

6. **Prioritize**
   - Sort by priority (critical > high > medium > low)
   - Within priority, sort by confidence

## Output Requirements

Return JSON:

```json
{
  "validated": [
    {
      "source": "agent-name",
      "type": "doc|automation|learning|followup",
      "suggestion": { /* original suggestion */ },
      "status": "new|duplicate|conflict|filtered",
      "reason": "Why this status",
      "mergedFrom": ["agent1", "agent2"] | null
    }
  ],
  "removed": [
    {
      "source": "agent-name",
      "suggestion": { /* removed suggestion */ },
      "reason": "Why removed"
    }
  ],
  "conflicts": [
    {
      "suggestion": { /* conflicting suggestion */ },
      "existingRule": "The existing rule text",
      "recommendation": "How to resolve"
    }
  ],
  "summary": {
    "total": 15,
    "new": 10,
    "duplicates": 3,
    "conflicts": 1,
    "filtered": 1
  }
}

## Example

### Phase 1 Input
```json
{
  "docAnalyzer": [
    { "content": "Use snake_case for hooks", "priority": "high" },
    { "content": "All API returns { data, error }", "priority": "high" }
  ],
  "learningExtractor": [
    { "title": "Hook naming convention", "description": "Hooks use snake_case" }
  ]
}

Existing Docs

## Code Style
- Hook files: snake_case (use_xxx.ts)

Output

{
  "validated": [
    {
      "source": "doc-analyzer",
      "type": "doc",
      "suggestion": { "content": "All API returns { data, error }", "priority": "high" },
      "status": "new",
      "reason": "Not documented, valuable pattern",
      "mergedFrom": null
    }
  ],
  "removed": [
    {
      "source": "doc-analyzer",
      "suggestion": { "content": "Use snake_case for hooks" },
      "reason": "Already documented in Code Style section"
    },
    {
      "source": "learning-extractor",
      "suggestion": { "title": "Hook naming convention" },
      "reason": "Duplicate of doc-analyzer suggestion, already documented"
    }
  ],
  "conflicts": [],
  "summary": {
    "total": 3,
    "new": 1,
    "duplicates": 2,
    "conflicts": 0,
    "filtered": 0
  }
}