Engineering
dev-cycle avatar

dev-cycle

Interactive development workflow manager. Coordinates discovery, planning, review, and build phases using a specialized team of AI agents (Scout, Bob, Garry, Arlo) for consistent project delivery.

Introduction

The dev-cycle skill provides a structured, multi-phase orchestration framework for software development within Claude Code. It moves beyond simple command execution by treating development as an interactive, session-based activity. The tool manages the complete lifecycle of a task—from initial codebase exploration to final implementation and validation—ensuring that requirements are clearly defined, edge cases are considered, and technical debt is minimized. By leveraging a team of specialized subagents, it distributes cognitive load and technical tasks to ensure both high-level alignment and implementation detail.

  • Orchestrates a four-phase workflow: Discovery (exploring requirements and existing patterns), Planning (defining actionable tasks and acceptance criteria), Review (aligning with the user), and Build (executing work through specialized agents).

  • Integrates specialized subagents: Scout for fast exploration, Bob for reliable technical construction, Garry for strategic oversight and coordination, and Arlo for complex data and analytics validation.

  • Tracks session state via persistent Markdown files stored in project-specific directories (active/completed), enabling seamless handoffs and session resumption.

  • Features a "Small Task Mode" to collapse overhead for trivial updates while still capturing documentation and decision logs.

  • Enforces branch-workflow integration, ensuring that code changes are committed and pushed following standard conventions.

  • Connects with megg for memory storage, ensuring context is loaded at the start of every cycle and key architectural decisions are preserved.

  • Intended for developers, software engineers, and technical leads who need a repeatable, auditable, and high-quality process for feature work or bug fixes.

  • Typical inputs include a high-level goal or feature request; the skill handles the granular breakdown, tool delegation, and progress reporting.

  • Constraints include dependencies on Node.js 20+ and the megg memory system; session files are stored in the filesystem, so it requires appropriate directory access.

  • Users should expect to interact with the system to approve plans or answer clarification questions during the discovery phase to maintain alignment.

  • The workflow is inherently iterative; tasks are only marked as complete when criteria such as test passage, linting, and security audits are met.

Repository Stats

Stars
14
Forks
1
Open Issues
1
Language
Shell
Default Branch
main
Sync Status
Idle
Last Synced
May 3, 2026, 09:26 AM
View on GitHub