extract-issues

name	extract-issues
description	Extract potential issues from note content with AI confidence tagging
feature_module	issues

Extract Issues Skill

Extract actionable issues from note canvas content using semantic analysis and confidence scoring.

Quick Start

Use this skill when:

• User finishes brainstorming in note canvas
• AI detects potential work items in note content
• User explicitly requests issue extraction (/extract-issues)

Example:

User writes in note:
"We need to implement JWT authentication and add rate limiting
to the API. Also fix the login bug where users can't log in with email."

AI extracts:
- Issue 1: "Implement JWT authentication" (RECOMMENDED)
- Issue 2: "Add API rate limiting" (RECOMMENDED)
- Issue 3: "Fix email login bug" (RECOMMENDED - explicit bug mention)

Workflow

1. Analyze Note Content

   • Parse note blocks (paragraphs, lists, headings)
   • Identify action items using verb patterns (implement, fix, add, create)
   • Detect explicit issue references ("bug", "feature", "task")

2. Extract Candidate Issues

   • For each potential issue:
     • Extract name (concise title, ≤100 chars)
     • Extract description (context from surrounding text)
     • Identify block_id for source linking
     • Detect labels from keywords (backend, frontend, security, etc.)
     • Infer priority from urgency markers (critical, urgent, nice-to-have)

3. Score Confidence

   • RECOMMENDED: Clear action item with explicit verb and context
     • Example: "Implement user authentication with JWT tokens"
   • DEFAULT: Implied action from discussion context
     • Example: "Authentication should use JWT" (implies implementation)
   • ALTERNATIVE: Possible interpretation, needs clarification
     • Example: "Maybe we should consider caching?" (tentative)

4. Return Structured Output

   • JSON array of issues with confidence tags
   • Preserve source block links
   • Include rationale for each confidence score

Output Format

{
  "issues": [
    {
      "name": "Implement JWT authentication",
      "description": "Add JWT-based authentication with refresh tokens to replace session cookies",
      "confidence": "RECOMMENDED",
      "source_block_id": "block-abc123",
      "labels": ["backend", "security"],
      "priority": "high",
      "rationale": "Clear implementation task with specific requirements (JWT, refresh tokens)"
    },
    {
      "name": "Add API rate limiting",
      "description": "Implement rate limiting middleware to prevent API abuse",
      "confidence": "RECOMMENDED",
      "source_block_id": "block-def456",
      "labels": ["backend", "infrastructure"],
      "priority": "medium",
      "rationale": "Explicit action item, standard security practice"
    },
    {
      "name": "Fix email login bug",
      "description": "Users cannot log in using email addresses, investigate validation logic",
      "confidence": "RECOMMENDED",
      "source_block_id": "block-ghi789",
      "labels": ["backend", "bug"],
      "priority": "critical",
      "rationale": "Explicit bug mention, blocks user access"
    }
  ],
  "summary": "Extracted 3 issues from note content",
  "total_blocks_analyzed": 8
}

Examples

Example 1: Feature Brainstorm

Input:

# New Dashboard Ideas

We should add a real-time activity feed showing recent updates.
Users need better filtering - by project, priority, and assignee.
Dark mode would be nice to have eventually.

Output:

{
  "issues": [
    {
      "name": "Add real-time activity feed",
      "confidence": "RECOMMENDED",
      "labels": ["frontend", "feature"],
      "priority": "medium",
      "rationale": "Clear feature request with specific requirement (real-time updates)"
    },
    {
      "name": "Implement advanced filtering",
      "confidence": "RECOMMENDED",
      "labels": ["frontend", "feature"],
      "priority": "high",
      "rationale": "User need explicitly stated with specific filters listed"
    },
    {
      "name": "Add dark mode theme",
      "confidence": "DEFAULT",
      "labels": ["frontend", "feature"],
      "priority": "low",
      "rationale": "Mentioned as 'nice to have', lower priority"
    }
  ]
}

Example 2: Bug Report

Input:

There's a critical bug in the login flow - users can't log in with email.
Investigation shows the regex validation is too strict.

Output:

{
  "issues": [
    {
      "name": "Fix email login validation bug",
      "confidence": "RECOMMENDED",
      "labels": ["backend", "bug"],
      "priority": "critical",
      "rationale": "Explicit critical bug with root cause identified"
    }
  ]
}

Example 3: Ambiguous Discussion

Input:

Maybe we should think about caching?
Could potentially improve performance.
Not sure if Redis or in-memory would be better.

Output:

{
  "issues": [
    {
      "name": "Evaluate caching strategy",
      "confidence": "ALTERNATIVE",
      "labels": ["backend", "performance"],
      "priority": "low",
      "rationale": "Tentative suggestion (maybe, could, not sure), needs discussion"
    }
  ]
}

Integration Points

• PilotSpaceAgent: Orchestrator routes to this skill via intent detection or /extract-issues command
• MCP Tools: Uses search_related_notes to find similar issues
• Approval Flow: Extracted issues require DEFAULT_REQUIRE_APPROVAL per DD-003
• Note Linking: Creates bidirectional links between issues and source blocks

References

• Design Decision: DD-013 (Note-First Workflow)
• Design Decision: DD-048 (Confidence Tagging)
• Schema: backend/src/pilot_space/ai/sdk/output_schemas.py