Purpose

Transform requirements, bug reports, review findings, or feature requests into well-structured, actionable tickets. Each ticket follows a consistent template with clear acceptance criteria, proper labels, priority classification, dependency links, and bounded scope so that an implementer agent or human developer can start work without ambiguity.

When to use

• A planning session has produced requirements that need to become trackable work items.
• A code review or security review has findings that need remediation tickets.
• A bug report needs to be formalised into a structured ticket.
• A large feature needs decomposition into individually assignable tickets.
• An audit identified gaps that need follow-up work items.

Do NOT use when

• Tickets already exist and need quality checking — use ticket-audit instead.
• The task is to implement what a ticket describes — use an implementer skill.
• The user wants to reorganise or reprioritise existing tickets — use backlog-verifier.
• The input is too vague to produce an actionable ticket — ask for clarification first.

Operating procedure

1. Parse the input source (requirements doc, review findings, bug description, or user message) and extract each distinct work item into a separate entry.
2. For each work item, determine the ticket type: bug, feature, chore, refactor, or spike.
3. Assign priority using this matrix: P0 = blocks release or causes data loss; P1 = significant user impact; P2 = moderate improvement; P3 = nice-to-have or cleanup.
4. Assign severity for bugs: Critical = system down; High = major feature broken; Medium = workaround exists; Low = cosmetic.
5. Write the ticket title as a concise imperative statement (e.g., "Add rate limiting to /api/login endpoint") — maximum 80 characters.
6. Write the Description section: one paragraph of context, the specific problem or need, and the expected outcome.
7. Write 3–5 Acceptance Criteria as checkboxes, each starting with a measurable verb: "Returns 429 status when rate exceeded", "Unit test covers the new path", "Documentation updated in API.md".
8. Assign labels from the project taxonomy: check for existing labels in .github/labels.yml, labels.json, or the ticket template, and reuse them.
9. Identify dependencies: search existing tickets for related work using grep -rl '<keyword>' tickets/ and add blocked-by: TICKET-ID or relates-to: TICKET-ID links.
10. Write the ticket to the appropriate location: create a markdown file in the tickets/ directory following the project naming convention, or output the ticket body for GitHub Issue creation.
11. Bound the scope explicitly: add a "Out of Scope" section listing adjacent work that is deliberately excluded from this ticket.

Decision rules

• One ticket = one deployable unit of work. If a ticket requires changes to 3+ unrelated systems, split it.
• Never create a ticket without at least 3 acceptance criteria.
• If the input is ambiguous, create a spike ticket to investigate rather than guessing requirements.
• Reuse existing labels — do not invent new ones unless no existing label fits.
• If a dependency target ticket does not exist yet, create it as a stub and note it needs fleshing out.

Output requirements

1. Ticket File(s) — one markdown file per ticket with these sections:
   • Title, Type, Priority, Severity (if bug), Labels
   • Description
   • Acceptance Criteria (checkbox list)
   • Dependencies (links)
   • Out of Scope
2. Creation Summary Table — | Ticket ID | Title | Type | Priority | Dependencies |
3. Dependency Map — text list showing which tickets block others.
4. Label Usage Report — labels applied and any new labels suggested.

References

• Project ticket templates in .github/ISSUE_TEMPLATE/ or docs/templates/
• Existing ticket index or manifest file
• Label taxonomy from .github/labels.yml or equivalent
• Input source document (requirements, review report, bug report)

Related skills

• ticket-audit — validates tickets created by this skill
• planner — produces the requirements that feed ticket creation
• security-review — generates findings that become remediation tickets
• team-leader — assigns created tickets to specialist agents

Failure handling

• If the input source is too vague to extract concrete work items, create a single spike ticket to clarify requirements.
• If the ticket directory or template does not exist, create the directory and note that a template should be established.
• If label taxonomy is missing, use a minimal default set: bug, feature, chore, high-priority, low-priority.
• If dependency targets cannot be verified, mark dependencies as unverified and flag for manual review.