| name | issue-triage |
| description | Triage a GitHub issue by classifying severity, labelling waste category, proposing next action, and drafting a structured response |
| compatibility | >=1.4 |
Issue Triage
Skill metadata: version "1.1"; license MIT; tags [triage, issues, github, lean, kaizen, bug, feature-request]; compatibility ">=1.4"; recommended tools [codebase, githubRepo, runCommands].
Systematically triage a GitHub issue using Lean waste categories, severity classification, and structured output — producing an actionable recommendation and a ready-to-post comment.
When to use
- The user says "triage this issue", "classify this bug", or "triage #NNN"
- A new issue has been opened and needs a first-response classification
- A batch of issues needs sorting before sprint planning
When NOT to use
- The issue is already triaged and has labels/milestone assigned — use Review Mode instead
- The user wants a full root-cause analysis (that is a debugging task, not triage)
Steps
-
Read the issue — Fetch the issue body, title, and all comments. Note the reporter, date, and any linked PRs or commits.
-
Reproduce context — Search the codebase for the affected file(s), function(s), or API mentioned in the issue. Read relevant code to confirm or deny the described behaviour.
-
Classify severity using these definitions:
critical — data loss, security vulnerability, crash in production path, or regression blocking release
major — significant functionality broken; workaround exists but is painful
minor — cosmetic, edge-case, or low-frequency impact
advisory — enhancement or improvement with no current breakage
-
Map to waste category (§6 Muda — use the full catalogue; most common for issues):
| Waste code | Name | Typical issue signal |
|---|
| W1 Overproduction | Duplicate feature/issue already in backlog | "This was reported in #NNN" |
| W2 Waiting | Blocked on external dependency, PR, or decision | "Waiting on upstream fix" |
| W5 Inventory | Stale issue; may no longer be reproducible | Reported > 6 months ago, no recent activity |
| W6 Motion | Poor UX causing user confusion — not a code defect | "I couldn't find the button" |
| W7 Defects | Code correctness bug | Crash, wrong output, data loss |
| W8 Unused talent | Missing automation or documentation causing repeated effort | "We keep getting this question" |
Codes W3, W4, W9–W16 may apply — consult §6 for the full definitions if the above don't fit.
-
Propose next action — choose one:
fix — assign to a sprint, link to affected file(s)
investigate — needs more information; list specific questions for the reporter
close-duplicate — link to the canonical issue
close-wontfix — explain the design decision
close-stale — the issue is no longer reproducible or relevant
defer — valid but low priority; move to backlog
-
Draft a comment — Write a structured GitHub comment using this template:
**Triage result**
| Field | Value |
|-------|-------|
| Severity | [critical / major / minor / advisory] |
| Waste category | [Wn — name] |
| Reproducible | [yes / no / unknown — explain] |
| Next action | [fix / investigate / close-duplicate / close-wontfix / close-stale / defer] |
**Summary**: [one sentence describing the issue and its impact]
**Recommended labels**: `[label1]`, `[label2]`
[If next action is investigate]:
> To help us investigate, please provide: [specific questions]
-
Verify — Confirm:
- Severity matches the impact described in step 2
- Waste category is justified by evidence from the codebase (not guessed)
- The drafted comment is factual and does not speculate beyond the code read
- No file content claimed without being read this session (§4 read-before-claiming)