CwF Features

Core Features

Structured Workflow Steps

CwF defines a consistent set of steps for every development task, split into planning and execution:

1. Task Planning: single-sentence goal, measurable success criteria, milestones, risk assessment
2. Requirements: functional requirements (FR), non-functional requirements (NFR), acceptance criteria
3. Design: architecture decisions, component design, data flow, interface contracts
4. Implementation Planning: files to modify, step-by-step implementation checklist
5. Implementation Execution: actual code written, deviations documented
6. Testing Planning: test strategy, test cases (Given/When/Then), validation criteria
7. Testing Execution: actual test results, pass/fail, lessons learned
8. Rollout: deployment plan, monitoring, rollback procedure
9. Maintenance: support notes, optimisation opportunities
10. Retrospective: variance analysis, lessons learned, process improvements

Not every task runs all 10 steps; CwF matches the workflow to the task type (see below).

Hierarchical Task Management

Tasks use decimal notation: 1, 1.2, 1.2.3. Unlimited nesting depth. Each level inherits from its parent. Parallel subtasks can be worked on independently.

Context Inheritance

Subtasks automatically inherit context from their parent tasks via token-efficient structural maps (~50–100 tokens per parent level vs 500–1000 for full files). When working on a subtask, the AI has access to the parent’s requirements, design decisions, and constraints; without you having to re-explain the project every session. This reduces context overhead by up to 80% in some steps of task execution.

CwF Skills

Skills are structured prompts that drive each workflow step. They tell the AI exactly what to read, what to write, and what to focus on at each stage. Skills are stored as files in your repo; versioned, reviewable, and customisable.

Decomposition Guidance

Every step includes a decomposition check with five universal signals (time, people, complexity, risk, independence). When 2+ signals trigger, CwF recommends breaking the task into subtasks. This prevents scope creep and keeps work manageable.

Retrospective Analysis

Post-completion variance tracking compares planned vs actual effort and outcomes. Lessons learned are captured in structured form and fed back into future task planning. Each completed task builds a record of how your team (or you and your AI) actually work.

Installation

See the full installation guide for prerequisites, the one-line install command, and post-install setup.

Supported Task Types