requesting-code-review

The requesting-code-review skill implements a standardized, high-discipline review process for coding agents. It automates the dispatch of the superpowers:code-reviewer subagent to analyze git commits between a base SHA and a head SHA. By isolating the reviewer from the main session history, this skill ensures objective, focused feedback on implemented features, bug fixes, and architectural changes. It is designed for engineers using subagent-driven development who want to catch bugs before they cascade into complex technical debt.

• Dispatches a focused subagent (code-reviewer) to verify implementation against specified project requirements.

• Facilitates systematic review cycles: mandatory for major features and merges, optional but recommended for debugging and refactoring.

• Provides a structured feedback mechanism: classifies issues into Critical, Important, and Minor, requiring resolution based on severity.

• Operates on Git-based commit ranges to ensure precise delta analysis between the working state and the baseline.

• Supports human-in-the-loop workflows by providing checkpoints for technical feedback and allowing push-back with reasoned arguments.

• Requires the definition of {WHAT_WAS_IMPLEMENTED}, {PLAN_OR_REQUIREMENTS}, {BASE_SHA}, and {HEAD_SHA} to generate context-rich reports.

• Follows a strictly non-negotiable policy for Critical issues; these must be resolved before proceeding to the next task in the development lifecycle.

• Encourages 'review early, review often' to maintain code health in long-running autonomous sessions.

• Integrates seamlessly with other Superpowers skills like subagent-driven-development and executing-plans.

• Ideal for use with tools like Claude Code, Cursor, and other CLI-based coding agents that manage Git workflows, as it prevents drift from the original technical design document.