HiAi-gg Triage GitHub Issues
Use this skill when an agent must work through an issue backlog: classify, prioritize, find duplicates, or recommend labels.
Workflow
1. Scope the set
- Define the query: repository, label/state filters, sort (e.g. oldest first, most commented, most recent).
- Fetch the issue list. Keep the working set bounded (e.g. top 50 by the chosen sort) and say what you are looking at.
2. Classify each issue
For each issue record:
- Type: bug / feature / docs / question / other (from title + body + labels).
- Severity signal: crash/data-loss wording, affected-version mentions, or "critical"/"urgent" in labels.
- Reproducibility: does the body include steps, environment, or logs?
- Actionability: is it self-contained, or does it need maintainer input?
3. Deduplicate
- Compare titles and body keywords across the set; group likely duplicates.
- For each candidate duplicate pair, state the evidence (same stack trace, same feature wording, same file references) — do not merge or close anything without explicit instruction.
4. Recommend labels and priorities
Propose labels (bug, enhancement, good-first-issue, triage-needed, …) and a priority ordering. Base priority on: user impact × affected-user count × maintainer effort. Label assignment is a write action — only apply labels when the user asks.
5. Report
Give a compact table or list: each issue (id + title), its classification, the recommended labels/priority, and any duplicate grouping. Separate "observed" from "recommended".