AgentSkillsCN

sprint

当项目已有规范或设计文档,且需要以原子级任务进行冲刺规划时,可使用此工具。

SKILL.md
--- frontmatter
name: sprint
description: Use when a project has a spec/design doc and needs sprint planning with atomic tasks

Sprint Planning

Break project specs into sprints and atomic tasks suitable for iterative development.

When to Use

  • Project has a spec, PRD, or design document
  • Need to plan implementation in sprints
  • Want atomic, commitable tasks with clear validation

Input

Provide the spec file path: @docs/spec.md, @DESIGN.md, etc.

Process

1. Analyze the Spec

Read the spec thoroughly. Identify:

  • Core features and requirements
  • Dependencies between components
  • Technical risks and unknowns
  • Integration points

2. Break Into Sprints

Each sprint must result in demoable software that:

  • Can be run and tested independently
  • Builds on previous sprints
  • Delivers visible progress

3. Define Atomic Tasks

Every task/ticket must be:

  • Atomic: Single commit, single concern
  • Testable: Has tests OR clear validation criteria
  • Independent: Can be worked on without blocking others (where possible)
  • Clear: Technical enough that implementation path is obvious

4. Task Format

markdown
## Task: [Short descriptive title]

**Depends on:** [task IDs or "none"]

### Description

[What needs to be built/changed]

### Acceptance Criteria

- [ ] [Specific, verifiable criterion]
- [ ] [Tests pass / validation method]

### Technical Notes

[Implementation hints, gotchas, relevant code paths]

5. Review with Subagent

After drafting, spawn a reviewer subagent:

code
Review this sprint plan for:
- Missing edge cases or requirements from the spec
- Tasks that are too large (should be split)
- Tasks that are too small (should be combined)
- Unclear acceptance criteria
- Missing dependencies
- Sprint goals that aren't demoable

Incorporate feedback, then write final plan.

6. Output Structure

Create two things:

Index file ai/SPRINTS.md:

markdown
# Sprint Plan: [Project Name]

Source: [spec file path]
Generated: [date]

| Sprint                        | Goal                | Status  |
| ----------------------------- | ------------------- | ------- |
| [01-auth](sprints/01-auth.md) | User authentication | active  |
| [02-api](sprints/02-api.md)   | Core API endpoints  | pending |
| [03-ui](sprints/03-ui.md)     | Frontend components | pending |

Individual sprint files ai/sprints/NN-name.md:

markdown
# Sprint 01: Auth

Goal: User authentication working end-to-end

## Tasks

### Task 1: Add JWT middleware

**Depends on:** none

**Description:** Add JWT validation middleware to /api routes

**Acceptance Criteria:**

- [ ] Middleware rejects invalid tokens
- [ ] Middleware extracts user ID
- [ ] Tests pass

### Task 2: ...

7. Updating Sprints

When a sprint completes:

  1. Update status in ai/SPRINTS.md index
  2. Mark tasks complete in the sprint file
  3. Set next sprint to active

Common Mistakes

MistakeFix
Tasks too vague ("implement auth")Be specific ("add JWT middleware to /api routes")
No validation criteriaEvery task needs tests or explicit verification
Sprints aren't demoableEach sprint = working software, not just "backend done"
Missing dependenciesMap what blocks what before finalizing order
Skipping reviewAlways have subagent review - catches blind spots