The /ticket Skill: Four Verbs to Drive Tickets to Resolution

Where this fits

You've just picked up a Jira ticket that reads "Fix login bug" — no repro, no expected behavior, no acceptance criteria. Instead of bouncing between chat, code, and the ticket comment thread, you type /ticket triage AUTH-123. The skill takes over: it reads the ticket, scans the repo, asks you scoped questions via /scope, and posts a locked work order directly as a ticket comment. That's the gap this skill fills — turning an unstructured ticket into a bounded, executable brief that an agent can drive from start to finish without re-asking for context.

Why agents benefit

• One verb per phase: triage, start, revise, finalize are full procedures in verbs/<verb>.md. The agent reads the procedure before acting, so behavior stays predictable across runs.
• Locked brief as source of truth: triage produces a work order as a ticket comment. All later steps reference that comment, so scope creep is visible and avoidable.
• Automatic sub-ordering: When a ticket is too large for a single agent context, the skill slices it into sub-orders in the same comment, keeping each agent focused and tracked.
• Statefulness via the ticket: Because the work order and its status live in the ticket, the agent can resume after a restart or hand off to another instance without losing context.
• No repo writes during triage: Triage only writes scope and spec into the ticket; code changes happen only when you explicitly start or revise — reducing accidental side effects.

Practical scenarios

Triage an ambiguous bug report

A user reports "App crashes on signup" with no stack trace. You run /ticket triage BUG-7. The agent reads the ticket, inspects the relevant code paths, and asks you for the missing device info. It then posts a work order with repro steps, expected behavior, and a suggested fix slice — all in one comment.

Start work from a locked brief

The work order from triage is approved. You run /ticket start PROJ-42. The agent reads that comment, creates a feature branch, implements the first slice, and pushes a draft PR for review. It tracks its own progress in the ticket thread.

Finalize a merged PR

After your team reviews and merges the PR, you run /ticket finalize PROJ-42. The agent verifies the merge, updates the ticket status, links the PR, and closes out the sub-orders. The ticket now has a complete audit trail.

Add it to your agent workflow

Install with:

npx skills add ConnorGriffin/skills --skill ticket

Then invoke it directly: /ticket start AUTH-42. The skill reads the ticket, checks the repo, and follows the procedure for the given verb. If you omit the verb or id, it asks you for them in one line — no dead ends.

Compare before adopting

Before you rely on this skill daily, compare it against your current process. Check last commit date on the GitHub repo and the issue tracker for maintenance signals. See if alternatives like a custom script or another agent skill cover the same lifecycle. Also verify it fits your ticket system (Jira, Linear, GitHub Issues) — the skill assumes a REST API and comment-based state.

Why it is worth tracking

With 20 stars and an active author (ConnorGriffin), this skill shows early but real momentum in the agent-skill ecosystem. It's not production-proven yet, but its design — explicit verbs, locked briefs, and sub-ordering — addresses a common failure mode: agents losing context across ticket transitions. Evaluate it on a low-stakes ticket first, and watch for updates to the repo and community feedback. For a directory view of this and similar skills, see https://www.openagentskill.com/skills/connorgriffin-ticket.