agentleFS
tomimortomimor/skills/skills/gh-issue-triage/SKILL.md

gh-issue-triage

Skill2 starsChanged 5 months agoMITView on GitHub

Scans the current GitHub repo for open, unassigned issues with no linked PR, scores them against an AI-readiness rubric, and returns the top 3 picks with rationale and a ready-to-paste agent kickoff prompt for each. Use when the user mentions AI triage, AI-ready issues, picking issues for an agent, what to work on next, or wants to find issues to dispatch to an AI.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 8 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run from the user's cwd:

    line 21
  2. Very specific:

    If that fails, tell the user the cwd is not a GitHub repo (or gh is not authenticated) and stop.

    line 27
  3. Very specific:

    If the result is empty, print No AI-ready issues in OWNER/REPO and stop.

    line 40
  4. Distinctive:

    If fewer than 3 issues clear the rubric, return as many as qualify.

    line 68
  5. Distinctive:

    After the last pick, print a one-line summary:

    line 112
  6. Distinctive:

    Do NOT invent acceptance criteria the issue does not state.

    line 120
  7. Distinctive:

    Do NOT pick issues with active reviewer debate, even if otherwise clear.

    line 122
  8. Distinctive:

    Do NOT pad the list to 3 when fewer issues qualify.

    line 123
  9. Distinctive:

    Do NOT include the full issue body in the kickoff prompt — distill it.

    line 124
  10. Distinctive:

    Do NOT recommend issues assigned to someone or already linked to a PR.

    line 125

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

GH Issue Triage

Find the top 3 GitHub issues in the current repo that are best suited for an AI agent to pick up, and produce a kickoff prompt for each.

This skill is read-only. Never assign, label, or comment on any issue — the output is for the user to act on.

Resolve the current repo

Run from the user's cwd:

gh repo view --json nameWithOwner -q .nameWithOwner

If that fails, tell the user the cwd is not a GitHub repo (or gh is not authenticated) and stop.

Fetch candidate issues

Pull open + unassigned issues that have no linked PR:

gh issue list --repo OWNER/REPO --state open --limit 50 \
  --search "no:assignee -linked:pr sort:updated-desc" \
  --json number,title,body,url,labels,comments,createdAt,updatedAt,author

If the result is empty, print No AI-ready issues in OWNER/REPO and stop.

Score each issue

Apply this rubric to every issue using its body, labels, and comment metadata. Each dimension scores 0–3 for a total of 0–15.

Dimension3 (great)0 (bad)
ClarityConcrete problem statement, no ambiguityVague or one-line title with no body
ScopeAcceptance criteria stated or strongly impliedOpen-ended "improve X" with no boundary
Self-containedNo business/product decisions or private context neededRequires stakeholder input or hidden context
SizeDoable in a single AI session"Rewrite the module" or cross-cutting refactor
Signalgood first issue / help wanted / ai-ready labels, low comment count, no debateHeated thread, conflicting requirements

Hard disqualifiers

Skip the issue (do not include in the ranking) if any of these are true:

  • Body or comments call for upfront design ("needs design", "needs spec", "TBD") or flag the issue as blocked on another task or external input.
  • Security-sensitive (auth, crypto, secrets handling) without clear scope.
  • Comment thread shows reviewers disagreeing on the desired behavior.
  • Title or labels indicate a question, discussion, or RFC — not work.

Pick the top 3

Sort by total score descending; break ties by updatedAt (most recent first). If fewer than 3 issues clear the rubric, return as many as qualify. Never pad the list to 3.

Output format

Render the headers, links, and bullets as plain markdown so the issue links stay clickable. Render the kickoff prompt itself inside a fenced code block so the user can copy it in one click and paste it into a fresh agent chat.

One block per pick, separated by ---:

## Top AI-ready issues in OWNER/REPO

### 1. [#NUMBER — Title](URL)
- Score: 13/15 (Clarity 3, Scope 3, Self-contained 2, Size 3, Signal 2)
- Labels: bug, good first issue
- Why it's AI-ready: <2-3 sentences distilled from the issue>

Kickoff prompt:

```
Implement issue #NUMBER in OWNER/REPO: <title>.

Context: <one-paragraph summary of the problem from the issue body>.

Acceptance criteria:
- <criterion 1>
- <criterion 2>

Out of scope: <anything explicitly deferred in the issue>.

Reference: <issue URL>
```

---

### 2. ...

The kickoff prompt is the deliverable — it must be self-contained enough to paste into a fresh agent chat (Cursor, Codex, Claude Code) and start work without further questions.

After the last pick, print a one-line summary:

Scanned N open issues in OWNER/REPO, M passed the rubric, top K shown.

Anti-patterns

  • Do NOT invent acceptance criteria the issue does not state. If the issue is light on detail, lower the Scope score; do not fabricate.
  • Do NOT pick issues with active reviewer debate, even if otherwise clear.
  • Do NOT pad the list to 3 when fewer issues qualify.
  • Do NOT include the full issue body in the kickoff prompt — distill it.
  • Do NOT recommend issues assigned to someone or already linked to a PR; the search query should exclude them, but double-check before ranking.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.