跳到正文
cccskills
免费在 GitHub 公开

split-into-tasks

Split an implementation-ready spec into the fewest independently deliverable vertical tasks and select only risk-justified intermediate review checkpoints. Use when a spec has multiple outcomes needing separate delivery or dependency tracking; keep one coherent outcome as the spec itself.

中文简介正在准备中。当前显示原文说明。

toy-crane/claude-code-playbook-template星标 77(GitHub 77 + cccskills 0)

实测评分 未实测

查看安装方法

包含的文件(4)

  • SKILL.md3.4 KB
  • agents/openai.yaml327 B
  • evals/evals.json15.0 KB
  • templates/task.md1.2 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Split Into Tasks

Read the spec

Require docs/specs/<slug>/spec.md; if it is missing, stop before proposing tasks.

Read the spec and current code, plus relevant project decisions, glossary terms, and an approved prototype when present. Treat missing implementation as the target gap. Pause on conflicting settled sources and present the exact decision to resolve. When the approved behavior forms one coherent user workflow, keep the spec as the sole handoff; separate operations or states inside that workflow do not create task boundaries.

Propose the complete outcome map

For work with multiple independently deliverable outcomes, draft the complete shallow map using the fewest independently usable and verifiable outcomes. Fold prerequisite work into the first outcome that makes it useful unless that work is independently deliverable. Declare only blockers that genuinely prevent a dependent outcome.

Cover every spec outcome. For each proposed task show its title, delivered behavior, observable acceptance criteria, focused deterministic verification, task-specific constraints, blockers with reasons, and any review checkpoint. Implementation derives the active outcome's technical approach just in time from the then-current repository, so omit predicted files, functions, code structure, technical-layer steps, context boundaries, and internal sequences. Under a task's constraints record only the references scoped to that task: the approved prototype screens and states it delivers, and a decision contract only it depends on. Name what each governs and leave its rules and screen compositions in the source itself.

Add an intermediate review checkpoint only when a material error could compound through substantial dependent work, or when deterministic checks cannot adequately settle a material risk such as security, data, permission, migration, recovery, or an external contract. Name its cumulative scope and concrete risk. A checkpoint buys one review pass over that scope, not a round that repeats until it reports nothing. The implementation phase owns the single final review over the whole diff.

Present the complete proposal and iterate until the user approves it. Write no task files before that approval.

Write the approved handoff

After approval, record shared settled constraints in the spec and only task-specific constraints in task files. Do not invent unsettled behavior.

Write each approved task to docs/specs/<slug>/tasks/<NN>-<slug>.md using templates/task.md, with blockers before dependents. For an approved revision, remove replaced tasks that have never recorded completion. Preserve a task with recorded completion history even if it later returned to in-progress or blocked; after its still-required obligations and blocker references move to the approved replacement, set it to superseded and append revision evidence naming the replacement and reason. A superseded task is terminal for that approved breakdown and is inactive recovery history, not part of the current delivery map. Do not create an archive. End after writing the current task handoff, before implementation.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Audit and establish current agent context for the technologies that define a project's stack. Use when setting up a project for agent work, after selecting or adding a framework, library, developer tool, or hosted service, or when entering an existing project whose agent context has not been audited.

中文简介正在准备中。当前显示原文说明。

toy-crane/claude-code-playbook-template772026年10月21日 更新

Revise active `docs/specs/SLUG/` specs so they match what has shipped since they were written. Use when earlier work has landed and queued specs may now rest on stale assumptions, acceptance criteria, constraints, terms, or prototypes, or when implementation reports a spec as stale before changing source. Takes one or more spec folders, or every folder under `docs/specs/` when none is named. Do not use to shape a new spec from a problem and direction, which belongs to `shape-idea`, or to reconcile durable project context and retire shipped folders, which belongs to `maintain-project-context`.

中文简介正在准备中。当前显示原文说明。

toy-crane/claude-code-playbook-template772026年10月21日 更新

Build a complete dummy-data HTML prototype or a temporary comparison of up to three alternatives for a specific user situation. Use to settle a screen-based product before implementation or resolve a visual or interaction decision during shaping or prototype review. Do not use for production implementation or CLI, terminal, or voice interfaces.

中文简介正在准备中。当前显示原文说明。

toy-crane/claude-code-playbook-template772026年10月21日 更新

commit

免费

Create Git commits for the current request's in-scope changes as logical Conventional Commits. Always use this skill for an actual Git commit operation, including requests to commit or amend changes, save work in repository history, stage and record selected files, or group staged or unstaged changes into commits. Use it even when the request describes recording changes without saying "commit"; it performs the commit rather than only suggesting a message, and it does not push.

中文简介正在准备中。当前显示原文说明。

toy-crane/claude-code-playbook-template772026年10月21日 更新

Interview the user to define or revise the app's overall product direction, then preserve the confirmed meaning in root PRODUCT.md. Always use for a new app direction or any app-wide change to users, usage situations, problem, promised change, core loop, durable boundaries, experience principles, success signals, assumptions, or unknowns, even when PRODUCT.md is not named. Start from an existing definition or a rough app direction. Do not use for blank-page idea discovery, architecture, or shaping or specifying one feature or work unit.

中文简介正在准备中。当前显示原文说明。

toy-crane/claude-code-playbook-template772026年10月21日 更新

Render an explanation as a diagram, comparison, trace with real values, or annotated code. Use when the user asks to have a concept, decision, flow, or code explained, or says they do not follow.

中文简介正在准备中。当前显示原文说明。

toy-crane/claude-code-playbook-template772026年10月21日 更新

查看 toy-crane 的全部技能

报告此技能的问题