Implementation Plan Skill

This skill provides guidelines for creating and managing implementation plans from design documents.

When to Apply

Apply this skill when:

•Translating design documents into actionable implementation plans
•Planning multi-session implementation work
•Breaking down large features into parallelizable subtasks
•Tracking implementation progress across sessions

Purpose

Implementation plans bridge the gap between design documents (what to build) and actual implementation (how to build). They provide:

•Clear deliverables with Rust type definitions
•Simple status tracking tables
•Checklist-based completion criteria
•Progress tracking across sessions

Plan Granularity

IMPORTANT: Implementation plans and spec files do NOT need 1:1 mapping.

Mapping	When to Use
1:N (one spec -> multiple plans)	Large specs should be split into smaller, focused units
N:1 (multiple specs -> one plan)	Related specs sharing dependencies can be combined
1:1 (one spec -> one plan)	Well-bounded features with clear scope

Recommended granularity:

•Each plan should be completable in 1-3 sessions
•Each plan should have 3-10 subtasks
•Maximize parallelizable subtasks

File Size Limits

CRITICAL: Large implementation plan files cause Claude Code OOM (Out of Memory) errors.

Hard Limits

Metric	Limit	Reason
Line count	MAX 400 lines	Prevents memory issues when agents read files
Modules per plan	MAX 8 modules	Keeps plans focused and manageable
Tasks per plan	MAX 10 tasks	Enables completion in 1-3 sessions

When to Split Plans

Split a plan into multiple files when ANY of these conditions are met:

•Line count exceeds 400 lines: Split by phase or module category
•More than 8 modules: Group related modules into separate plans
•More than 10 tasks: Break into logical sub-plans
•Multiple phases with dependencies: Create separate plans per phase

Splitting Strategy

code

BEFORE (one large plan):
impl-plans/active/foundation-and-core.md (1100+ lines)

AFTER (split by phase):
impl-plans/active/foundation-interfaces.md (~200 lines)
impl-plans/active/foundation-mocks.md (~150 lines)
impl-plans/active/foundation-types.md (~150 lines)
impl-plans/active/foundation-core-services.md (~200 lines)

Split Plan Naming Convention

When splitting, use consistent naming:

•{feature}-{phase}.md - For phase-based splits
•{feature}-{category}.md - For category-based splits

Example:

•session-groups-types.md
•session-groups-repository.md
•session-groups-manager.md

Cross-References Between Split Plans

Each split plan MUST include:

markdown

## Related Plans
- **Previous**: `impl-plans/active/foundation-interfaces.md` (Phase 1)
- **Next**: `impl-plans/active/foundation-core-services.md` (Phase 3)
- **Depends On**: `foundation-interfaces.md`, `foundation-types.md`

Output Location

IMPORTANT: All implementation plans MUST be stored under impl-plans/ subdirectories.

code

impl-plans/
├── README.md              # Index of all implementation plans
├── active/                # Currently active implementation plans
│   └── <feature>.md       # One file per feature being implemented
├── completed/             # Completed implementation plans (archive)
│   └── <feature>.md       # Completed plans for reference
└── templates/             # Plan templates
    └── plan-template.md   # Standard plan template

Directory Rules

Directory	Purpose
`impl-plans/active/`	Implementation plans currently in progress
`impl-plans/completed/`	Archived completed plans for reference
`impl-plans/templates/`	Plan templates and examples

DO NOT create implementation plan files outside impl-plans/.

Implementation Plan Structure

Each implementation plan file MUST include:

1. Header Section

markdown

# <Feature Name> Implementation Plan

**Status**: Planning | Ready | In Progress | Completed
**Design Reference**: design-docs/<file>.md#<section>
**Created**: YYYY-MM-DD
**Last Updated**: YYYY-MM-DD

2. Design Document Reference

•Link to specific design document section
•Summary of what is being implemented
•Scope boundaries (what is NOT included)

3. Modules and Types

List each module with its Rust type definitions. USE ACTUAL RUST CODE for traits, structs, enums, and function signatures - not prose descriptions.

markdown

## Modules

### 1. Core Traits

#### src/interfaces/filesystem.rs

**Status**: NOT_STARTED

```rust
pub trait FileSystem: Send + Sync {
    fn read_file(&self, path: &Path) -> Result<Vec<u8>>;
    fn write_file(&self, path: &Path, content: &[u8]) -> Result<()>;
    fn exists(&self, path: &Path) -> Result<bool>;
    fn watch(&self, path: &Path) -> Result<impl Stream<Item = WatchEvent>>;
}

#[derive(Debug, Clone)]
pub struct WatchEvent {
    pub event_type: WatchEventType,
    pub path: PathBuf,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum WatchEventType {
    Create,
    Modify,
    Delete,
}

Checklist:

code


### 4. Status Tracking Table

Use simple tables for overview tracking:

```markdown
## Module Status

| Module | File Path | Status | Tests |
|--------|-----------|--------|-------|
| FileSystem trait | `src/interfaces/filesystem.rs` | NOT_STARTED | - |
| ProcessManager trait | `src/interfaces/process.rs` | NOT_STARTED | - |
| Mock implementations | `src/test/mocks/*.rs` | NOT_STARTED | - |

5. Dependencies

Simple table showing what depends on what:

markdown

## Dependencies

| Feature | Depends On | Status |
|---------|------------|--------|
| Phase 2: Repository | Phase 1: Traits | BLOCKED |
| Phase 3: Core Services | Phase 1, Phase 2 | BLOCKED |

6. Completion Criteria

Simple checklist:

markdown

## Completion Criteria

- [ ] All modules implemented
- [ ] All tests passing
- [ ] cargo build passes
- [ ] cargo clippy passes
- [ ] Integration verified

7. Progress Log

Track session-by-session progress:

markdown

## Progress Log

### Session: YYYY-MM-DD HH:MM
**Tasks Completed**: Module 1, Module 2
**Tasks In Progress**: Module 3
**Blockers**: None
**Notes**: Discovered edge case in variable parsing

Content Guidelines

INCLUDE Rust Code

ALWAYS include actual Rust code for:

•Trait definitions
•Struct definitions
•Enum definitions
•Type aliases
•Function signatures (name, parameters, return types)

Example:

markdown

```rust
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct SessionGroup {
    /// Format: YYYYMMDD-HHMMSS-{slug}
    pub id: String,
    pub name: String,
    pub status: GroupStatus,
    pub sessions: Vec<GroupSession>,
    pub config: GroupConfig,
    pub created_at: DateTime<Utc>,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum GroupStatus {
    Created,
    Running,
    Paused,
    Completed,
    Failed,
}

code


### DO NOT Include

- Implementation logic (function bodies)
- Private functions
- Algorithm details
- Excessive prose descriptions

### Format Comparison

**GOOD** (Rust-first):
```markdown
#### src/interfaces/clock.rs

```rust
pub trait Clock: Send + Sync {
    fn now(&self) -> DateTime<Utc>;
    fn timestamp(&self) -> String;
    async fn sleep(&self, duration: Duration);
}

Checklist:

• Define Clock trait
• Export from lib.rs

code


**BAD** (Prose-heavy):
```markdown
**Exports**:
| Name | Type | Purpose | Called By |
|------|------|---------|-----------|
| `Clock` | trait | Time operations | Caching, logging |

**Function Signatures**:
now() -> DateTime<Utc>
  Purpose: Get current date/time
  Called by: Logger, Cache

Parallelization Rules

Subtasks can be parallelized when:

•No data dependencies between tasks
•No shared file modifications
•No order-dependent side effects

Mark dependencies explicitly in the status table.

Workflow Integration

Creating a Plan

•Read the design document
•Identify feature boundaries
•Define Rust traits, structs, and enums
•List modules with status tracking
•Set completion criteria
•Create plan file in impl-plans/active/

During Implementation

•Update module status as work progresses
•Add progress log entries per session
•Note blockers and decisions
•Check off completion criteria

Completing a Plan

•Verify all completion criteria met
•Update status to Completed
•Add final progress log entry
•Move file to impl-plans/completed/

Quick Reference

Section	Required	Format
Header	Yes	Markdown metadata
Design Reference	Yes	Link + summary
Modules	Yes	Rust code blocks + checklist
Status Table	Yes	Simple table
Dependencies	Yes	Simple table
Completion Criteria	Yes	Checklist
Progress Log	Yes	Session entries