AGENT SKILL

create-tickets

Breaks a sprint architecture into sequenced, numbered implementation tickets with dependency ordering

License 许可
MIT
最近更新
2026-09-25
Stars
★ 0
Forks
0

01 前言 Front Matter

name
create-tickets
description
Breaks a sprint architecture into sequenced, numbered implementation tickets with dependency ordering

03 SKILL.md 正文

Create Tickets Skill

This skill breaks a sprint's architecture document into actionable implementation tickets. It is typically invoked by the sprint-planner agent during planning, but can also be used standalone.

Inputs

  • Sprint's sprint.md (must exist) -- its Architecture and Use Cases sections, sized to the sprint's effort decision (may read "N/A -- trivial" for a trivial/small sprint)

Process

  1. Read artifacts: Read the sprint's sprint.md Architecture and Use Cases sections.

  2. Identify work units: Break the Sprint Changes into coherent implementation units. Each unit should be completable in one focused session.

  3. Order by dependency: Number tickets so foundation work comes before features. Record dependencies in each ticket's depends-on field.

  4. Create ticket files: Use the create_ticket MCP tool. Each ticket gets tickets/NNN-slug.md with:

    • YAML frontmatter: id, title, status (open), use-cases, depends-on
    • Description and acceptance criteria (checkboxes)
    • Implementation plan: approach, files to create/modify, testing plan, documentation updates

    Issue lifecycle: When you call create_ticket(sprint_id, title, issue=<filename>), the referenced issue file is physically moved from clasi/issues/ into <sprint>/issues/ and its frontmatter is updated to status: in-progress. When all tickets referencing that issue are moved to done, Issue.move_to_done() is called automatically, which moves the file into <sprint>/issues/done/. No manual move_issue_to_done call is needed in the happy path.

    Multi-issue sprints — pass issue= explicitly: create_ticket only auto-links a ticket to the sprint's issue when the sprint has exactly one linked issue. On any sprint linked to 2+ issues, omit- ting issue= leaves the ticket's issue: field empty (it does not guess by linking every sprint issue). Pass issue=<filename> explicitly on every create_ticket call for a multi-issue sprint.

    Multi-ticket issue propagation: When multiple tickets implement the same source issue, every ticket must carry the issue: back-reference. Use create_ticket(issue=filename) for the first ticket. For subsequent tickets, call add_issue_ref(ticket_path, issue_filename) after creation. Before returning from ticket creation, verify that every ticket working toward an issue has a non-empty issue: field.

  5. Propagate references: Copy TODO and GitHub issue references to ticket frontmatter. List GitHub issues in the sprint doc's ## GitHub Issues section.

  6. Verify coverage: Every use case covered by at least one ticket. Every ticket traces to at least one use case.

  7. Verify sequencing: No circular dependencies. Foundation before features.

Ticket Format

File: <sprint-dir>/tickets/NNN-slug.md

---
id: "NNN"
title: Short title
status: open
use-cases: [SUC-001, SUC-002]
depends-on: [NNN]
---

Followed by: description, acceptance criteria, and implementation plan.

Output

Numbered ticket files in the sprint's tickets/ directory, ready for implementation.

04 同仓库更多 Skills