do

The "do" skill is an orchestration-focused AI skill designed to execute phased implementation plans through coordinated subagents. Rather than performing implementation work directly, the skill acts as a centralized orchestrator that delegates tasks to specialized subagents responsible for implementation, verification, quality review, anti-pattern detection, and commit management. Its primary purpose is to ensure structured, traceable, and verifiable execution of complex plans while reducing the risk of undocumented assumptions, skipped validation steps, or inconsistent workflows. The skill is especially suited for executing plans that were previously defined in a separate planning phase.

A core capability of the skill is its strict execution protocol. Each phase of work is assigned to dedicated subagents with clearly defined objectives and evidence requirements such as commands executed, outputs generated, and files modified. The orchestration flow enforces verification before progression, requiring post-phase validation through verification, anti-pattern review, and code quality review subagents before any commit activity is allowed. The skill also emphasizes documentation-driven implementation by instructing implementation subagents to copy existing documented patterns, verify APIs instead of assuming capabilities, and reference documentation sources when working with unfamiliar interfaces. Between phases, branch synchronization and handoff coordination are managed through additional subagents to preserve continuity while allowing fresh execution contexts.

This skill is intended for software engineering workflows where phased execution, validation, and coordination are critical. It can be used by development teams, AI-assisted coding environments, technical project coordinators, or engineers managing large implementation plans that require structured oversight. Typical use cases include executing multi-phase feature implementations, coordinating verification-heavy development processes, enforcing coding standards during automated execution, and managing workflows where strict validation and review gates are required before code integration or commits.