agentleFS
y-miyazakiy-miyazaki/config/.agents/skills/github-issue-triage/SKILL.md

github-issue-triage

Skill1 starsChanged 8 days agoApache-2.0View on GitHub

Classify GitHub Issues, apply allowlisted triage labels, and post analysis or clarifying questions as Issue comments. Use in conversation when asked to triage an Issue (URL or number) without opening a PR. Also the L1 loop entry skill for Issue events. Default is API side effects (labels/comments) in both modes; never edits repository files.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When marking triage:ready, comment that a human may add the autofix label to request a draft PR — do not invoke autofix.

    line 63

“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “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

UTILITY SKILL — Issue triage (labels + comments), not autofix or PR creation.

Input

  • Automation: detect JSON in the prompt with issue facts from detect_issue.sh. agent_edit_allowed in ## Constraints is ignored by this skill (see Edit gate). When the prompt names a message directory (a line Loop message directory: <path>), the run is a message run: write files there instead of posting.
  • Interactive: natural-language Issue URL or number; there is no detect JSON. Read the Issue with read-only gh issue view <n> --json number,title,body,state,labels,author,comments; same label/comment rules.

Output Specification

Session report per common-output-format.md. Side effects are GitHub Issue labels and comments, not repository file edits. Interactive: applied via gh and scripts/issue_comment.sh. Message run: message.md (comment markdown) and optional message_labels.json ({"add":[...],"remove":[...]}) in the message directory; the platform posts the comment and applies the labels.

Edit gate (exception)

Unlike other loop entry skills, the Interactive default is not survey-only: triaging an Issue means applying labels and posting comments, so those GitHub API side effects run in both modes. This skill never edits repository files; agent_edit_allowed and agent_edit_target do not change its behavior, whether true or false. A user who wants a dry run must say so; then report the planned labels and comment without applying them.

Execution Scope

USE FOR:

  • Classify Issue type when confident (bug, feature, question, documentation)
  • Drive label FSM (needs-triage, triage:needs-info, triage:ready, triage:failed) per category-fsm.md
  • Post analysis and clarifying questions as Issue comments

DO NOT USE FOR:

  • Opening or revising PRs (github-issue-autofix / github-pr-revise)
  • Firing repository_dispatch or starting github-issue-autofix (autofix starts only from a human action, such as adding the autofix label)
  • Mentions / @ triggers
  • Applying labels outside scripts/labels.json
  • Relying on AskUserQuestion — questions go to the Issue thread

Reference Files Guide

  • category-fsm.md (always read)
  • category-prompt-rules.md (always read)
  • common-checklist.md (always read)
  • common-output-format.md (always read)
  • category-pr-body-links.md (read when writing an Issue comment)

Workflow

  1. Automation: parse detect JSON (result.issue_number, title, body, labels, event). Interactive: resolve the Issue from the user's URL or number with gh issue view; use its current labels as the FSM input state.
  2. Automation: skip silently if detect already skipped (bot / policy). Interactive: if the Issue carries triage:failed, report that a human must remove it first and stop.
  3. Load allowlist from scripts/labels.json. Interactive only: ensure missing allowlisted labels exist before apply (gh label create only for catalog entries). Message run: do not create labels.
  4. Classify when confident; if low confidence → keep/apply needs-triage only and ask clarifying questions on the Issue (no question self-censorship — see category-prompt-rules.md).
  5. Pick the FSM event per category-fsm.md § Choosing the event (reopened → opened; a human reply while triage:needs-info → re-classify, then mark_ready or mark_needs_info) and compute the transition. Interactive: apply them via gh issue edit / gh api. Message run: do not run gh issue edit or any label write; write the transition (label_fsm_next_state output {"add":[...],"remove":[...]}) to message_labels.json in the message directory. Type labels (bug, feature, question, documentation) are outside the FSM helper: Interactive — add them with gh issue edit <n> --add-label <type> (and --remove-label for a replaced type); message run — append them to the add list (and a replaced type to remove) in the same message_labels.json.
  6. Comment per category-prompt-rules.md. Interactive: post via scripts/issue_comment.sh — create on each FSM transition or re-triage; correct only for minor edits to the latest marked bot comment. Message run: post nothing; write the comment markdown (no preamble, status line, or leading ---) to message.md in the message directory — one comment per run.
  7. When marking triage:ready, comment that a human may add the autofix label to request a draft PR — do not invoke autofix.
  8. Emit session report. Never modify repository files.

Error Handling

ConditionSeverityAction
Detect status: errorFatalStop; do not mutate Issue
Interactive: Issue carries triage:failedInfoReport that a human must remove triage:failed first; do not change labels or comment
Unknown / non-allowlisted label requestedRecoverableDo not apply; keep needs-triage
gh failureRecoverableReport in session output; avoid partial unlabeled state when possible
Unsafe partial failureRecoverableApply allowlisted triage:failed (message run: add it to message_labels.json); detect skips until a human removes it

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.