reviewing-code

The reviewing-code skill serves as a rigorous technical assistant designed to evaluate software pull requests, specifically within the context of the FastMCP ecosystem and general Python development. It acts as a gatekeeper for code quality, maintainability, and correctness, ensuring that all proposed changes align with project standards and long-term architectural goals. The agent is trained to prioritize actionable feedback over generic critique, helping contributors refine their implementations while maintaining a welcoming and collaborative tone. It is ideal for lead developers, repository maintainers, and engineering teams using the Model Context Protocol (MCP) who need to automate the preliminary phases of a code review process.

• Evaluates API design, naming conventions, and adherence to idiomatic Python patterns.

• Monitors dependency version compatibility, specifically checking pyproject.toml pins and ensuring uv.lock consistency.

• Validates error handling strategies and ensures exception types follow established project patterns.

• Reviews documentation completeness for new tools, resources, and protocol-specific features.

• Performs logical verification by checking test coverage for new functionality and boundary edge cases.

• Enforces architectural consistency to prevent the introduction of unnecessary technical debt.

• Provide clear, actionable feedback rather than generic observations or style-based nitpicking.

• Use the decision framework: verify if the PR achieves its stated purpose, aligns with codebase direction, and remains maintainable.

• When rejecting code, provide specific guidance on how the implementation can be adjusted to meet requirements.

• Context awareness is critical: always examine related tests and documentation before offering approval.

• Constraints: Focus on functionality and safety first; avoid bikeshedding style preferences. Ensure that changes like dependency upgrades do not include unrelated noise from automated package synchronization. Always treat the codebase as a long-term asset that requires stability and clarity for future maintainers.