create-implementation-plan（制定实施计划）

Implementation Plan Creator

Creates implementation plans that define how to execute a project — phases, tasks, sequencing, testing, and milestones. Does not define system architecture. References a technical design when one exists.

Core Principle: Vertical Slices

Structure every plan as vertical slices following the tracer bullet strategy. Each phase delivers a thin but complete slice through all layers.

• Phase 1 is always the tracer bullet: the simplest possible end-to-end flow that proves the architecture works
• Every phase produces a deployable increment that can be demonstrated and validated
• Testing is embedded in each phase — never deferred to a standalone phase
• No phase is purely "setup" or "infrastructure" without user-visible functionality

Task ID Convention

Task IDs are derived from the feature slug and are unique across the entire plan. Format: PREFIX-PN-NN

• PREFIX — feature slug uppercased and shortened to 3–5 characters (e.g. user-onboarding → ONBD, payment-refunds → PAY)
• PN — phase number (e.g. P1, P2)
• NN — sequential task number within the phase (e.g. 01, 02)

Example: ONBD-P1-01, ONBD-P1-02, ONBD-P2-01

Phase Structure

Each phase must define:

• Goal: User-facing or system-visible capability this phase delivers
• Vertical slice: Which layers are touched and what flows end-to-end
• Tasks: Specific work items with owners and estimates
• Testing: What is tested in this phase and how
• Acceptance criteria: How to verify the phase goal
• Dependencies: What must be complete before this phase starts

Example Phase Progression

• Phase 1 (Tracer Bullet): Single entity, one endpoint, basic persistence, one happy-path test — proves the full stack end-to-end
• Phase 2: Core business logic, error handling, validation — primary use case complete
• Phase 3: External integrations, async processing — system connects to the outside world
• Phase 4: Edge cases, advanced features, performance — production hardening
• Phase N (Final): Phased rollout, monitoring verification, documentation — production readiness

Routing

• Define requirements → /create-prd
• Design architecture → /create-technical-design
• Write the code → /implement

Process

Step 1: Locate Upstream Artifacts

Check .specs/[feature-slug]/ for existing documents. If the feature slug is not yet known, ask for the feature name first and derive the slug.

• Technical design found (.specs/[feature-slug]/TECHNICAL-DESIGN.md): Read it. Use it as the source of truth for architecture, components, APIs, and constraints. Reference it explicitly throughout the plan. Do not re-derive architectural decisions — cite the design.
• PRD found but no technical design (.specs/[feature-slug]/PRODUCT-REQUIREMENTS.md): Read it. Then do a quick research phase: scan the relevant codebase to understand existing patterns, integration points, and constraints well enough to derive the architectural context needed for the plan.
• Neither found: Do a quick research phase — scan the codebase for existing patterns, identify relevant entry points and integration boundaries, and derive scope, approach, and constraints from the task description and code. Keep this focused and efficient; the goal is just enough context to write a solid plan.

Step 2: Gather Context

Ask for (skip anything already answered by upstream artifacts):

• Project/feature name and team members
• Target timeline or milestone dates
• Any known constraints (hard deadlines, team availability, external dependencies)

Step 3: Collect Phase Information

Ensure you have enough to structure phases with vertical slices:

• What is the simplest end-to-end flow that proves the architecture?
• What is the core business logic that must be complete for V1?
• Are there external integrations or async processes?
• What does production readiness require (monitoring, rollback, docs)?

Step 4: Determine Critical Sections

Testing Strategy: test types (unit, integration, e2e, contract, load) with scope, approach, and critical scenarios per type. Each phase embeds its own testing — this section describes the overall strategy and test data management approach.

Rollback Plan: deployment strategy (feature flags, phased rollout percentages), rollback triggers with specific thresholds, rollback steps (immediate, database, communication), post-rollback process (RCA, fix, re-test, re-deploy).

Step 5: Generate and Validate

File location: .specs/[feature-slug]/IMPLEMENTATION-PLAN.md — derive the slug from the feature name (lowercase, hyphen-separated). Create the directory if it does not exist.

Generate using the template below. Validate against the checklist. Highlight any gaps in the output summary.

Template

# Implementation Plan — [Project/Feature Name]

| Field        | Value                        |
| ------------ | ---------------------------- |
| Tech Lead    | @Name                        |
| Team         | Name1, Name2, Name3          |
| Epic/Ticket  | [Link]                       |
| Status       | Draft / In Review / Approved |
| Created      | YYYY-MM-DD                   |
| Last Updated | YYYY-MM-DD                   |

## Overview

Brief description of what is being built and the execution strategy.

**Technical Design**: [link to TECHNICAL-DESIGN.md if available, or "Not provided"]

## Implementation Phases

Structured as vertical slices following tracer bullet strategy. Testing is embedded in every phase.

### Phase 1: [Tracer Bullet — simplest end-to-end flow]

**Goal**: [User-facing or system-visible capability this phase delivers]

**Vertical slice**: [Which layers are touched and what flows end-to-end]

**Tasks**:

| ID | Task | Owner | Estimate |
|----|------|-------|----------|
| [PREFIX-P1-01] | [Task 1] | @Name | [estimate] |
| [PREFIX-P1-02] | [Task 2] | @Name | [estimate] |

**Testing**:

- [Test type]: [Specific scenarios verified in this phase]

**Acceptance Criteria**:

- [ ] [Verifiable outcome 1]
- [ ] [Verifiable outcome 2]

**Dependencies**: [What must be complete before this phase starts]

---

### Phase 2: [Core business logic complete]

[Same structure as Phase 1]

---

### Phase N: [Production readiness]

[Same structure as Phase 1]

## Milestones

| Milestone | Target Date | Description |
|-----------|-------------|-------------|
| [Name] | YYYY-MM-DD | [What is delivered at this point] |

## Dependencies

| Dependency | Type | Owner | Status | Risk if Delayed |
|------------|------|-------|--------|-----------------|
| [Item] | External / Internal / Technical | @Owner | [Status] | [Impact] |

## Risks

| Risk | Impact | Probability | Mitigation |
|------|--------|-------------|------------|
| [Description] | H/M/L | H/M/L | [Strategy] |

Critical sections (Testing Strategy, Rollback Plan) follow after Risks when applicable.

Validation Checklist

• Technical design referenced (if one exists)
• Phase 1 is a tracer bullet delivering end-to-end value
• Every phase has goal, tasks, testing, acceptance criteria, and dependencies
• All task IDs follow the PREFIX-PN-NN convention and are unique across the plan
• Testing is embedded in each phase — no standalone test phase
• No phase is purely setup or infrastructure without user-visible output
• Milestones are defined with target dates
• Dependencies listed with owners and delay risk
• Risks has 3+ entries with impact, probability, and mitigation
• Testing Strategy present (all projects)
• Rollback Plan present (production deployments)

Output Summary

After generating, provide:

Implementation Plan Created: "[Project Name]"
File: .specs/[feature-slug]/IMPLEMENTATION-PLAN.md

Phases: N
Estimated Total Duration: [sum of estimates]

Technical Design: [Linked / Not provided]

Sections Included:
- Mandatory: [list]
- Critical (N/N): [list]

Missing (if any):
- [Section]: [why it matters]

Suggested Next Steps:
- Review phases with team
- Break tasks into tickets/issues
- Run /implement to begin execution