ftw_skill

Overview

This skill is a PIV (Plan-Implement-Validate) workflow orchestrator that runs a systematic multi-phase development loop for feature work. It manages PRD discovery/creation, PRP generation, codebase analysis, multi-agent execution, validation, debug loops, and smart commits. Use it to enforce repeatable, phase-by-phase delivery with automated validation and clear escalation paths.

How this skill works

The orchestrator parses input to detect either a direct PRD path or a project path and switches between execute and discovery modes. In discovery mode it runs an interactive questionnaire, proposes a PRD and project phases, writes PRD and setup files, then transitions into the phase workflow. For each phase it spawns fresh sub-agents to do codebase analysis, generate PRPs, execute changes, validate results, and run up to three debug iterations before committing or escalating.

When to use it

• Building features incrementally with clearly defined phases and acceptance criteria
• Coordinating multi-agent workflows where each step needs fresh context
• Generating PRDs and PRPs from an initial discovery conversation
• Automating validation and debug loops to reduce manual oversight
• Running repeatable, auditable phase-based releases with semantic commits

Best practices

• Always ensure a PRD exists or run discovery so phases are scoped before execution
• Provide as much initial context as possible to skip redundant discovery questions
• Keep phases focused and small to minimize debug iterations
• Treat sub-agents as stateless workers—pass only the necessary context and artifacts
• Review validator reports before allowing automatic commits; escalate when human input is requested

Example use cases

• Create a PRD from a short product brief, auto-generate PRPs, and run phase 1 through validation
• Iteratively implement a multi-step feature (API → backend → frontend) with automated tests and validation in each phase
• Archive and reproduce previous archived project workflows by replaying phase executions
• Run a safety net for CI by spawning executors and validators that independently verify changes before merges