通过分诊角色驱动的状态机分诊 issues。适用于用户想创建 issue、分诊 issues、审查 incoming bugs 或 feature requests、为 AFK agent 准备 issues,或管理 issue workflow 时。
Triaging issues in a project's issue tracker through a small state machine of triage roles. Authored in Chinese, it defines two category roles, bug and enhancement, and five state roles, needs-triage, needs-info, ready-for-agent, ready-for-human, and wontfix, with each triaged issue carrying exactly one category and one state role. A firm rule requires every comment or issue posted during triage to begin with a disclaimer line noting that the content was generated by AI during triage.
The maintainer invokes it with a /triage command and a natural-language request, such as showing everything that needs attention or moving a specific issue to a given state. For attention triage it queries the tracker and presents three buckets, unlabeled, needs-triage, and needs-info with reporter activity, oldest first. Triaging a specific issue proceeds through gathering context, recommending a category and state with rationale, reproducing the problem for bugs, optionally grilling the reporter for detail, and applying the outcome.
Outcomes are concrete: ready-for-agent posts an agent brief so an unattended (AFK) agent can pick the issue up, ready-for-human uses the same structure but explains why the work cannot be delegated, needs-info posts structured triage notes from a template, and wontfix either closes with an explanation or records an enhancement in an out-of-scope knowledge base. Reference docs for agent briefs and the out-of-scope knowledge base, plus a resume flow for previously triaged issues, round it out.
Two category roles, bug and enhancement, and five state roles, needs-triage, needs-info, ready-for-agent, ready-for-human, and wontfix. Each triaged issue carries exactly one of each.
Yes. Every comment or issue posted during triage must begin with a disclaimer line stating that the content was generated by AI during triage.
The maintainer runs the /triage command and describes what they want in natural language, for example moving issue 42 to ready-for-agent or asking what needs attention.
Before grilling, it attempts to reproduce the problem by reading the reporter's steps, tracing relevant code, and running tests, then reports success, failure, or insufficient detail.
An agent brief comment, following the agent-brief reference, so an unattended AFK agent can pick up and implement the issue.
Quick Setup:
.claude/skills/Repository
vinvcn/mattpocock-skills-zh-cn