Issue Triage Playbook

Use this Skill when an agent needs to inspect a GitHub issue queue and turn it into a useful next-step recommendation.

When to use this Skill

Use it when:

• a repository has multiple open issues and the user wants prioritization help
• the agent should gather the latest issue facts before suggesting work
• a draft follow-up comment is needed before a human approves a write

Do not use this Skill to silently post comments or close issues without an explicit instruction and approval step.

Triage process

1. Read the relevant issues first.
2. Identify which issues are:
   • blocking users
   • newly regressed
   • missing reproduction detail
   • obviously stale
3. Separate:
   • facts from assumptions
   • draft wording from actions that should actually be executed
4. End with a clear recommendation:
   • address now
   • ask for more detail
   • defer
   • close

Preferred output

For each issue the agent surfaces, include:

• issue number and short title
• why it matters now
• evidence from the issue body or recent comments
• recommended next step

If the user asks for a comment draft, prepare the draft but do not execute the write unless explicitly told to do so.

Execution

Read or update issues through the helper script:

./scripts/run.sh '{"action":"list_issues","owner":"agentpm-dev","repo":"agentpm-examples","state":"open"}'

The helper also accepts:

• ./scripts/run.sh --input-file payload.json
• ./scripts/run.sh to fall through to plain agentpm run @zack/github-issues

See:

• references/tool-contract.md
• references/triage-decision-cues.md
• references/examples.md

Failure cases to avoid

• Ranking a bug as urgent without citing impact or recency
• Suggesting closure when the issue still has an untested reproduction
• Posting a follow-up comment when the user only asked for a draft
• Treating a maintainer question as resolved work