recommend-assignee

name	recommend-assignee
description	Recommend assignee based on expertise matching and workload balance
feature_module	["issues","members"]

Recommend Assignee Skill

Suggest optimal assignee for issue using expertise matching, past work analysis, and workload balancing.

Quick Start

Use this skill when:

• New issue created without assignee
• User requests assignee recommendation (/recommend-assignee)
• Issue reassignment needed due to skill mismatch

Example:

Issue: "Implement GraphQL API for user queries"

AI recommends:
- [e-mail] (RECOMMENDED)
  - Expertise: GraphQL, API design (15 prior issues)
  - Workload: 3 open issues (below team average of 5)
  - Recent: Completed "GraphQL schema design" 2 weeks ago

Workflow

1. Analyze Issue Requirements

   • Extract technical keywords from title/description
   • Identify domain (backend, frontend, infra, design)
   • Classify complexity (low, medium, high)

2. Build Team Expertise Profile

   • For each workspace member:
     • Past issues worked on (completed, in-progress)
     • Labels/technologies from those issues
     • Success rate (completion time, quality)
   • Load expertise data from user_expertise table

3. Score Candidates

   • Expertise Match (0-100): Keyword overlap with past work
   • Workload (0-100): Inverse of current open issues
   • Recency (0-100): Recently worked on similar issues
   • Domain Familiarity (0-100): Experience in issue domain
   • Weighted score: Expertise (40%) + Workload (30%) + Recency (20%) + Domain (10%)

4. Apply Business Rules

   • Exclude members on PTO or unavailable
   • Consider role permissions (e.g., only admins for security issues)
   • Respect team preferences (preferred domains)

5. Tag Confidence

   • RECOMMENDED: High score (>80), clear expertise match
   • DEFAULT: Medium score (50-80), general fit
   • ALTERNATIVE: Multiple candidates with similar scores

Output Format

{
  "recommendations": [
    {
      "user_id": "user-abc123",
      "email": "[e-mail]",
      "name": "Alice Johnson",
      "confidence": "RECOMMENDED",
      "score": 92,
      "rationale": "Strong GraphQL expertise (15 prior issues), below workload average",
      "expertise_match": {
        "graphql": 95,
        "api_design": 88,
        "python": 82
      },
      "workload": {
        "open_issues": 3,
        "team_average": 5
      },
      "recent_work": [
        {"title": "GraphQL schema design", "completed": "2024-01-10"}
      ]
    },
    {
      "user_id": "user-def456",
      "email": "[e-mail]",
      "name": "Bob Smith",
      "confidence": "DEFAULT",
      "score": 68,
      "rationale": "API experience but no GraphQL history, workload is high",
      "expertise_match": {
        "api_design": 75,
        "python": 88
      },
      "workload": {
        "open_issues": 7,
        "team_average": 5
      }
    }
  ],
  "summary": "2 candidates found, 1 RECOMMENDED"
}

Examples

Example 1: Backend Issue

Input:

{
  "issue": {
    "title": "Optimize database query performance",
    "description": "User list query takes 5+ seconds with 10k users",
    "labels": ["backend", "database", "performance"]
  }
}

Output:

{
  "recommendations": [
    {
      "email": "[e-mail]",
      "confidence": "RECOMMENDED",
      "score": 94,
      "rationale": "Database expert (PostgreSQL indexing), completed 8 perf issues",
      "expertise_match": {"postgresql": 98, "performance": 92, "backend": 85}
    }
  ]
}

Example 2: Frontend Issue

Input:

{
  "issue": {
    "title": "Implement responsive dashboard layout",
    "labels": ["frontend", "ui", "react"]
  }
}

Output:

{
  "recommendations": [
    {
      "email": "[e-mail]",
      "confidence": "RECOMMENDED",
      "score": 88,
      "rationale": "Frontend specialist, recently completed similar dashboard work",
      "expertise_match": {"react": 95, "ui": 90, "css": 85},
      "recent_work": [
        {"title": "Dashboard redesign", "completed": "2024-01-05"}
      ]
    },
    {
      "email": "[e-mail]",
      "confidence": "ALTERNATIVE",
      "score": 86,
      "rationale": "Similar expertise, slightly higher workload",
      "expertise_match": {"react": 92, "ui": 88},
      "workload": {"open_issues": 6, "team_average": 5}
    }
  ]
}

Example 3: Security Issue

Input:

{
  "issue": {
    "title": "Fix XSS vulnerability in comment form",
    "labels": ["security", "bug", "frontend"]
  }
}

Output:

{
  "recommendations": [
    {
      "email": "[e-mail]",
      "confidence": "RECOMMENDED",
      "score": 96,
      "rationale": "Security specialist with XSS remediation experience",
      "expertise_match": {"security": 98, "xss": 95, "frontend": 80},
      "certifications": ["OWASP Top 10 Training"]
    }
  ]
}

Integration Points

• AssigneeRecommenderAgent: Primary agent implementing this workflow
• MCP Tools: Uses get_workspace_members, search_issues
• Data Sources: user_expertise table (updated on issue completion)
• Approval Flow: Recommendations are AUTO_EXECUTE (suggestions only)

References

• Design Decision: DD-048 (Confidence Tagging)
• Agent: backend/src/pilot_space/ai/agents/assignee_recommender_agent_sdk.py
• Database: user_expertise table schema