← ClaudeAtlas

devpilot-auto-featurelisted

End-to-end workflow for implementing OpenSpec changes with TDD and mandatory code review checkpoints between task chapters. Use instead of openspec-apply-change when the user wants quality gates, TDD, reviews between chapters, or a full implement-review-archive-sync cycle. Wraps openspec-apply-change with superpowers:test-driven-development per task, superpowers:requesting-code-review after each chapter, plus automatic openspec archival, spec syncing, and CLAUDE.md/README.md update checks. Triggers on "auto feature", "implement with tdd", "implement with reviews", "build this end to end", "implement [change] with quality gates", "full workflow", "tdd and review", or any request to implement an OpenSpec change mentioning testing, reviews, quality, or thoroughness. Also triggers when resuming a partially-completed change. If the user just says "implement [change]" without quality keywords, use openspec-apply-change instead.
SiyuQian/devpilot · ★ 4 · AI & Automation · score 70
Install: claude install-skill SiyuQian/devpilot
# Auto Feature Implement OpenSpec changes end-to-end using test-driven development, with code review checkpoints after each chapter of tasks. This skill orchestrates multiple superpowers skills into a disciplined workflow that produces high-quality, reviewed code. The workflow: for every task, write tests first, then implement. After finishing each chapter (a group of tasks sharing a major number — e.g., 1.1, 1.2, 1.3), pause and request a code review. Address all reasonable feedback before moving on. Once everything is done, archive the change, sync specs, and check if documentation needs updating. ## Why this workflow matters Implementing features without review gates leads to compounding issues — a small design misstep in chapter 1 becomes a painful refactor in chapter 3. By reviewing after each chapter, problems surface early when they're cheap to fix. TDD ensures each task has verification built in from the start, so reviews can focus on design and intent rather than "does it work?" ## Workflow ### Phase 1: Setup 1. **Select the change** — Follow the same selection logic as `openspec-apply-change`: - If a name is provided, use it - Auto-select if only one active change exists - If ambiguous, run `openspec list --json` and ask the user to pick 2. **Load context** — Run `openspec instructions apply --change "<name>" --json` and read all context files (proposal, specs, design, tasks). Understand the full picture before touching code. 3. **Parse task struct