reviewing-code

This skill acts as a specialized software agent reviewer for FastMCP and Model Context Protocol (MCP) implementations. It assists developers in maintaining high-quality code standards, ensuring that contributions align with the project's evolution and architectural philosophy. The agent rigorously evaluates pull requests not only for correctness but for maintainability and strategic alignment with the FastMCP framework. It provides actionable feedback by analyzing code logic, dependency pins, and API design, ensuring that changes do not introduce regressions or technical debt. It is ideal for lead maintainers and contributors working within the FastMCP ecosystem, ranging from server development and client integration to AuthKit and OAuth authentication examples.

• Performs deep analysis of dependency version compatibility, specifically monitoring pyproject.toml and uv.lock integrity.

• Validates API design patterns, flagging mutable default arguments, ambiguous naming, and non-idiomatic Python constructs.

• Evaluates functionality against stated PR purpose, enforcing a 'burden of proof' model where contributors must clearly demonstrate value.

• Reviews implementation details related to MCP-compliant resources, tools, and prompts, ensuring adherence to the official protocol SDK standards.

• Provides constructive, actionable feedback aimed at improving code ergonomics and long-term project sustainability.

• Focus on functional correctness: Before approving, verify that code handles error conditions, edge cases, and required development workflow steps (e.g., pytest, uv sync).

• Enforce backward compatibility: Ensure that changes to dependency versions or API definitions do not silently break existing users; demand explicit handling for version shifts.

• Maintain stylistic consistency: Discourage nitpicking on subjective style, prioritizing instead the project's established conventions, naming patterns, and user-centric API design.

• Strategic alignment: Always reject unwanted features regardless of code quality if they do not advance the codebase toward its intended future direction.

• Inputs expected: Pull request diffs, related documentation snippets, and surrounding test coverage context.

• Outputs provided: Specific, non-generic recommendations for refactoring, architectural adjustments, and necessary verification steps for merge readiness.