---

name: triage description: Auto-triage issues by classifying type, priority, and suggesting labels version: 1.0.0 author: xiaoju111a triggers:

• triage
• classify
• categorize

---

Issue Triage Skill

Automatically classify and triage GitHub issues to help maintainers quickly process incoming issues.

Purpose

• Classify issue type (bug, feature, question, docs, etc.)
• Assess priority based on impact and urgency
• Suggest appropriate labels from repository's label set
• Provide actionable recommendations for maintainers

Issue Type Classification

Priority Assessment

Priority Signals

Critical/High Priority Signals

• Security-related keywords: "vulnerability", "CVE", "security", "exploit", "injection"
• Data issues: "data loss", "corruption", "deleted", "missing data"
• Availability: "down", "unavailable", "500 error", "timeout"
• Scale: "all users", "everyone", "production"

Low Priority Signals

• Cosmetic: "typo", "formatting", "color", "alignment"
• Edge cases: "rare", "edge case", "specific scenario"
• Nice-to-have: "would be nice", "minor", "small improvement"

Analysis Guidelines

1. Read the full issue - Title and body together provide context
2. Check for reproduction steps - Well-documented bugs are easier to assess
3. Consider the author - First-time contributors may phrase things differently
4. Look at linked issues/PRs - May provide additional context
5. Assess impact scope - How many users are affected?

Codebase Analysis

This skill includes a scan_codebase script that automatically:

1. Extracts keywords from issue title and body
2. Searches the codebase for relevant files
3. Returns code snippets for context

How It Works

• Identifies function names, file paths, and technical terms in the issue
• Searches source files (.py, .js, .ts, .go, etc.) for matches
• Provides relevant code context to improve classification accuracy

Benefits

• Confirms if mentioned features/files exist in codebase
• Helps identify which components are affected
• Increases confidence in classification

Output Format

Copy{
    "type": "bug|feature|enhancement|question|documentation|other",
    "priority": "critical|high|medium|low",
    "labels": ["label1", "label2", "label3"],
    "confidence": "high|medium|low",
    "summary": "One-line summary of the issue",
    "reason": "Brief explanation of classification reasoning",
    "related_files": ["src/auth.py", "src/login.js"]
}

Rules

1. Be conservative - Only suggest labels you're confident about
2. Maximum 4 labels - Don't over-label issues
3. Use repo labels - Only suggest labels that exist in the repository
4. Acknowledge uncertainty - If unsure, set confidence to "low" and explain
5. Consider context - Repository type matters (library vs app vs docs)

Special Cases

Needs More Information

If the issue lacks critical details:

• Set confidence to "low"
• Note what information is missing in the reason
• Consider suggesting "needs-info" or "needs-triage" label if available

Duplicates

If the issue appears to be a duplicate:

• Note in the reason
• Suggest "duplicate" label if available
• Don't assign other type labels

Multi-Type Issues

If an issue contains both a bug report and feature request:

• Classify based on the primary concern
• Note the secondary aspect in the reason
• Recommend splitting if appropriate