Issue Triage

Fetch GitHub issues, apply intelligent analysis, and visualize priorities.

When to Use

• Starting a work session and need to decide what to tackle
• Triaging a backlog with many competing priorities
• Looking for issues that can be delegated to AI coding agents
• Identifying urgent vs. important vs. quick-win issues

Quick Start

# 1. Fetch issues (saves to triage-data.json)
./scripts/fetch-issues.sh owner/repo

# 2. Launch dashboard in browser
./scripts/serve.sh

How This Skill Works

This skill combines deterministic data fetching with LLM judgment:

1. Script fetches issues → outputs structured triage-data.json
2. You analyze each issue using the scoring criteria below
3. Update the JSON with scores and analysis
4. Use the viewer to sort, filter, and explore prioritized issues

The script handles data retrieval; you provide the intelligence that only an LLM can offer.

Data Format

The fetch script outputs JSON in this structure:

{
  "metadata": {
    "repository": "owner/repo",
    "generated": "2025-12-27T01:00:00Z",
    "total_issues": 42
  },
  "issues": [
    {
      "number": 123,
      "title": "Issue title",
      "body": "Full issue body...",
      "body_preview": "First 500 chars...",
      "labels": [{"name": "bug", "color": "d73a4a"}],
      "label_names": ["bug"],
      "age_days": 7,
      "days_since_update": 2,
      "comment_count": 5,
      "is_assigned": false,
      "url": "https://github.com/...",
      "scores": {
        "delegation": null,
        "importance": null,
        "urgency": null,
        "clarity": null,
        "effort": null,
        "priority": null
      },
      "analysis": null
    }
  ]
}

Scoring Criteria

For each issue, evaluate these criteria and assign scores (1-5):

🤖 Delegation Potential

Can this be delegated to an AI coding agent like Copilot?

🎯 Importance

How important is this to the project's success?

⚡ Urgency

How time-sensitive is this?

🧹 Clarity

How well-defined is the issue?

⏱️ Effort Estimate

How much work is this likely to be?

Priority Formula

Priority = (Importance × 3) + (Urgency × 2) + (Clarity × 1.5) + (Delegation × 1) + (Effort × 0.5)

Max score: 40 | High priority: ≥30 | Medium: 20-29 | Low: <20

Workflow

Standard Workflow

1. Run ./scripts/fetch-issues.sh owner/repo to fetch issues
2. Analyze the top issues and provide a summary with priorities
3. Run ./scripts/serve.sh to launch the dashboard in the browser
4. The dashboard auto-loads triage-data.json and displays interactive visualizations

Manual Exploration

1. Run ./scripts/fetch-issues.sh owner/repo
2. Run ./scripts/serve.sh (opens http://localhost:8080/dashboard.html)
3. Use the dashboard to explore, filter, and drill into issues
4. Press Ctrl+C in terminal to stop the server when done

Viewer Features

The viewer.html provides:

• Sort by priority, age, comments, delegation score
• Filter by label, scored/unscored status
• Search issues by title or number
• Score issues with click-to-edit interface
• Export your scored data as JSON
• Dark mode GitHub-style interface

Example LLM Output

When analyzing issues, output updates in this format:

{
  "number": 123,
  "scores": {
    "delegation": 5,
    "importance": 2,
    "urgency": 2,
    "clarity": 5,
    "effort": 5,
    "priority": 24.5
  },
  "analysis": "Trivial docs fix. Perfect for AI delegation - exact change specified."
}

Or provide a summary report alongside the JSON updates.

Files

issue-triage/
├── SKILL.md             # This file
├── dashboard.html       # Full analytics dashboard
├── viewer.html          # Simple issue viewer
├── scripts/
│   ├── fetch-issues.sh  # Data fetching script
│   └── serve.sh         # Local server + browser launch
└── triage-data.json     # Generated data (git-ignored)

Notes

• The viewer works offline - all processing is client-side
• Drag and drop JSON files onto the viewer to load them
• Scores persist in the JSON; export to save your work
• For private repos, authenticate gh CLI first