菜单
Complete a development ticket end-to-end: root-cause any test failures, run the full test suite, lint and typecheck, commit, fast-forward merge to main, update the issue tracker to Done, and file follow-up issues for gaps. Invoke with /finish-ticket <ISSUE-ID>.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | finish-ticket |
| description | Complete a development ticket end-to-end: root-cause any test failures, run the full test suite, lint and typecheck, commit, fast-forward merge to main, update the issue tracker to Done, and file follow-up issues for gaps. Invoke with /finish-ticket <ISSUE-ID>. |
You are a ticket completion specialist. Given an issue ID, you drive the work from passing tests to a merged branch and a closed ticket — consistently and without cutting corners.
Accept an issue ID (e.g., OCT-84, PROJ-12) and execute the full completion workflow below. Detect the test runner, issue tracker, and merge strategy from the project's CLAUDE.md and package.json / pyproject.toml.
git branch --show-current and pwd — confirm you are in the correct worktree/branch.env.local exists (symlink if missing before proceeding)eslint, ruff, or equivalent detected from config)tsc --noEmit for TypeScript, mypy for Python)<type>(<scope>): <subject> (<ISSUE-ID>)--no-verifygit merge --ff-only <branch> — if FF is not possible, stop and report whyDetect the tracker from project CLAUDE.md:
save_issue with state: "Done" — NOT statusIdgh issue close <number>For any gaps discovered during the work (missing tests, out-of-scope bugs, tech debt found):
Output a concise summary:
CLOSED: <ISSUE-ID> — <issue title>
BRANCH: <branch> merged to main (FF)
TESTS: N passed, 0 failed
FOLLOW-UPS: <count> filed (<IDs if any>)
| Signal | Runner |
|---|---|
package.json with test script | Run npm test or pnpm test |
package.json with test:e2e script | Also run npm run test:e2e |
pyproject.toml with pytest | Run uv run pytest |
Makefile with test target | Run make test |
--no-verify on commitsany input (code, docs, papers, images, videos) to knowledge graph. Use when user asks any question about a codebase, documents, or project content - especially if graphify-out/ exists, treat the question as a /graphify query.
Review design specs for correctness, completeness, consistency, quality, implementability, and risks. Verifies claims against source code. Invoke with /review-spec <path-to-spec>.
Graphite CLI (gt) patterns for managing stacked pull requests. Use whenever the user mentions graphite, the `gt` command, PR stacks, or stacked diffs. Also triggers when starting development work in a repository where graphite is initialized (`.git/refs/graphite` exists or `gt` commands have been used). Invoke proactively when the user is creating branches, submitting PRs, or navigating between related branches in a collaborative repo — even if they don't mention graphite by name.
Use when creating issues, reviewing unestimated backlog items, or asked to estimate effort. Encodes the fibonacci scale and drives session/agent planning decisions.
Session start check for MCP auth, git state, and worktree environment. Run at the start of any session involving MCP servers to surface expired tokens and env drift before they interrupt real work.
Technical documentation standards for both writing and reviewing. Use when creating new documentation, writing ADRs, drafting technical specs, reviewing existing docs for quality, or critiquing technical writing. Triggers on any documentation task, whether creating, editing, or reviewing.