github-issues-triage

name	github-issues-triage
description	Triage a repository's open issue backlog by proposing priorities, labels, and comments. Detects already-triaged issues, identifies potential duplicates, and generates a gh-based apply script.

GitHub Issues Triage

Purpose

Use this skill to triage the backlog of open GitHub issues for a given repository. The script will:

• Export all open issues (with pagination), available labels, and issue templates
• Propose priority (critical/high/medium/low) based on content analysis
• Propose labels to apply to each issue
• Propose a gentle comment when an issue does not adhere to the issue templates
• Skip already-triaged issues (those with substantive labels)
• Detect potential duplicates based on title/content similarity
• Produce:
  • A markdown report with proposed labels/priorities/duplicates
  • A runnable script (using gh) to apply the labels and comments

Requirements

• gh CLI installed and authenticated (gh auth login)
• Repository coordinates in OWNER/REPO format
• Python 3.8+

Workflow

Step 1: Export Repository Data

Run the triage script to export issues, labels, and templates:

python3 scripts/triage_issues.py --repo OWNER/REPO --out ./triage-output

See scripts/triage_issues.py for the export script.

Outputs:

File	Description
issues.json	All open issues (PRs filtered out)
labels.json	All repository labels
templates/	Downloaded issue templates
proposals.json	Proposed labels/comments per issue
report.md	Human-readable triage report
apply.sh	Executable script to apply proposals

Step 2: Review and Refine Proposals

1. Review labels.json to understand the project's labeling taxonomy
2. Review templates under templates/ to determine adherence criteria
3. Review each issue in issues.json and refine proposals.json

Guidelines:

• Prefer existing labels; do not invent new ones unless explicitly asked
• Be conservative: if uncertain, leave proposed_labels empty and note why
• If an issue does not follow templates, propose a respectful comment

Step 3: Apply Labels and Comments

After reviewing report.md and proposals.json:

bash ./triage-output/apply.sh

Key Features

Priority Detection

The script analyzes issue title, body, and labels to propose a priority:

Priority	Keywords detected
🔴 Critical	security, vulnerability, CVE, production down, blocking, urgent
🟠 High	regression, crash, bug, error, broken, important
🟡 Medium	enhancement, feature request, improvement
🟢 Low	typo, documentation, cosmetic, minor

Already-Triaged Detection

Issues are skipped if they have substantive labels (not just triage-related labels like "needs-triage" or "needs-info"). This prevents re-triaging issues that have already been processed.

Duplicate Detection

The script compares issue titles and bodies using text similarity (threshold: 70%). Potential duplicates are flagged in the report for manual review.

Example Output

Sample proposals.json

[
  {
    "number": 42,
    "url": "https://github.com/octocat/hello-world/issues/42",
    "title": "App crashes on startup",
    "proposed_labels": ["needs-triage", "high"],
    "proposed_priority": "high",
    "proposed_comment": "Thanks for opening this issue...",
    "reasons": ["Issue has no labels", "Detected priority: high"],
    "already_triaged": false,
    "potential_duplicates": [
      {"number": 38, "title": "Application crash at launch", "similarity": 0.75, "match_type": "title"}
    ]
  }
]

Sample report.md

The report includes:

• Summary: Total issues, needs triage count, already triaged count
• Priority breakdown: Count per priority level
• Issues needing triage: Table sorted by priority
• Potential duplicates: Table showing similar issues
• Already triaged: Collapsed list of skipped issues

Sample apply.sh

Only includes issues that need triage (skips already-triaged):

#!/usr/bin/env bash
set -euo pipefail
REPO='octocat/hello-world'

gh issue edit -R "$REPO" 42 --add-label 'needs-triage' --add-label 'high'
gh issue comment -R "$REPO" 42 --body 'Thanks for opening this issue...'

References

• Technical Reference - Detailed script documentation
• Examples - Additional usage examples
• Report Template - Template for report output
• Proposals Schema - JSON schema for proposals