/eiffel.tasks - Phase 3: Task Decomposition

Purpose: Break contracts into implementable chunks with clear acceptance criteria.

Usage

/eiffel.tasks <project-path>

Example:

/eiffel.tasks d:\prod\simple_cache

Project Scoping

All files are created inside the PROJECT directory:

<project-path>/
├── .eiffel-workflow/
│   ├── tasks.md                       (implementation tasks)
│   ├── prompts/
│   │   └── phase3-task-review.md      (optional AI review)
│   └── evidence/
│       └── phase3-tasks.txt

Prerequisites

• Phase 2 complete: <project-path>/.eiffel-workflow/synopsis.md exists and approved

Workflow

Step 1: Read Inputs

Read:

• <project-path>/src/*.e (contracts)
• <project-path>/.eiffel-workflow/approach.md (implementation strategy)
• <project-path>/.eiffel-workflow/synopsis.md (resolved issues)

Step 2: Generate Tasks

Create <project-path>/.eiffel-workflow/tasks.md:

# Implementation Tasks: [Library Name]

## Task 1: [Feature/Component Name]
**Files:** src/[file].e
**Features:** [feature_name], [feature_name]

### Acceptance Criteria
- [ ] Preconditions satisfied
- [ ] Postconditions verified by contracts
- [ ] Compiles clean
- [ ] Skeletal tests pass

### Implementation Notes
[From approach.md - relevant algorithm details]

### Dependencies
[Other tasks that must complete first]

---

## Task 2: [Next Feature]
...

Step 3: Generate Optional AI Review Prompt

Create <project-path>/.eiffel-workflow/prompts/phase3-task-review.md:

# Task Completeness Review

Review the task breakdown for completeness.

## Check For:
- Features in contracts not covered by tasks
- Missing dependencies between tasks
- Tasks too large (should be split)
- Tasks too small (should be combined)

## Tasks
[PASTE tasks.md]

## Contracts
[PASTE src/*.e]

## Output
List any gaps or issues found.

Step 4: Human Approval

Present task list and ask:

"[N] tasks identified. Review the task breakdown. Approve to proceed to Phase 4 (Implementation)?"

DO NOT PROCEED until user explicitly approves.

Step 5: Save Evidence

Save to <project-path>/.eiffel-workflow/evidence/phase3-tasks.txt:

# Phase 3 Task Evidence
# Project: <project-path>
# Date: [timestamp]

Tasks created: [count]
Dependencies identified: [count]
Human approved: [yes/no]

# Status: APPROVED

Context Management (RLM Pattern)

This skill focuses ONLY on: <project-path>

DO NOT:

• Read files outside this project directory
• Load entire ecosystem or multiple libraries into context
• Keep large file contents in working memory

DO:

• Use Task tool with Explore agent for ecosystem questions
• Ask targeted questions when task dependencies involve external libraries
• Release context after getting answers

Example - Ecosystem Query:

Task depends on simple_mml patterns? Don't read all its files.
Instead: Task(subagent_type=Explore) → "How does simple_mml handle set operations?"

The sub-agent searches, summarizes, and returns ONLY what you need. Your context stays focused on this project.

Completion

Phase 3 COMPLETE: Tasks defined.
Project: <project-path>

Task file: .eiffel-workflow/tasks.md
Task count: [N]

Next: Run /eiffel.implement <project-path> to write feature bodies.