Skip to main content

🕵️ Code Review Mode

A comprehensive AI-powered reviewer that performs multi-dimensional analysis of your codebase.

Overview​

The Code Review mode performs an in-depth, context-aware review across six critical dimensions: Compliance, Code Quality, Readability, Linting, Correctness, and Completeness. It goes beyond syntax checks to ensure the code aligns with architectural expectations and regulatory requirements using project-level documents like the PRD and TRD, with first-class support for Godspeed projects.

Primary Use Cases​

  • Performing a full audit of project code.
  • Ensuring feature completeness and alignment with specifications.
  • Detecting compliance or security gaps (e.g., GDPR, logging policies).
  • Validating code readability and documentation quality.
  • Providing inline comments and a structured Markdown review report.
  • Applying framework-specific best practices for Godspeed projects.

Example Prompts​

"Review this Godspeed project for correctness and completeness."
"Check this PR for compliance issues."
"Evaluate if this function matches the spec in the TRD."
"Run a full review excluding node_modules and .env."

Internal Capabilities / Agents​

  • Compliance Checker: Identifies policy violations (e.g., GDPR, logging).
  • Code Quality Analyzer: Reviews maintainability, structure, and design.
  • Readability Assessor: Evaluates naming conventions, comments, and logical flow.
  • Linting Integrator: Checks formatting against coding guidelines.
  • Correctness Validator: Looks for logic bugs and edge cases.
  • Completeness Verifier: Matches features against requirements in PRD/TRD.
  • Godspeed Best Practices Enforcer: Applies framework-specific validations via RAG MCP.

Model Behavior & Tool Access​

CapabilityStatus
Read files✅ (project files, PRD.md, TRD.md, .gitignore)
Write files✅ (review reports in .md + inline comments)
Execute commands✅ (within secure scope for lint/scan triggers)
Evaluate code✅
Use external documentation✅ (via RAG MCP)
Modify source code❌ (suggestions only, no direct edits)

Preferred Models:​

Claude 3 Opus / GPT-4.5 (Holistic analysis, complex reasoning) Claude 3 Sonnet (Context parsing, inline comments)

Scope​

✅ In Scope

  • Source code review in most languages (JS/TS/Go/Python/C++).
  • Review of Markdown documentation (PRD, TRD, README).
  • Analysis of Godspeed configuration files.
  • Review of Git diffs, pull requests, and staged changes.
  • Generation of structured review reports and inline comments.

❌ Out of Scope

  • UI/UX design critiques (visual layer).
  • Review of proprietary binary formats.
  • Analysis of external services not under version control.
  • Direct modification of source code (only suggestions are provided).
  • Live deployment or CI/CD execution.

Embedded Behaviors​

  • Reads PRD and TRD to align code review with project requirements and architecture.
  • Performs a six-dimensional review (Compliance, Code Quality, Readability, Linting, Correctness, Completeness).
  • Supports interactive setup to select specific review focus areas.
  • Automatically fetches Godspeed documentation via MCP for relevant validations.
  • Provides Godspeed-aligned feedback for event handlers, workflows, and datasources.
  • Automatically generates structured and explainable test coverage comments for untested Godspeed flows.

🔄 Workflow Integration Stores review reports in:

reviews/
  └── review-YYYY-MM-DD-HHMM.md

Recommended documentation structure:

docs/
  ├── PRD.md
  └── TRD.md

Leverages:​

  • rag-node MCP to fetch Godspeed documentation and best practices.
  • command group for optional linting or scanning triggers.
  • ask mode to explain review results or clarify issues.
  • coach mode for converting review findings into learning goals.

Safety Features​

  • Provides suggestions only; does not modify source code directly.
  • Inline comments are non-destructive.
  • Command execution (e.g., linting, scanning) is sandboxed.
  • Reviews are fully reproducible via prompts and report logs.

Shortcut to Activate​

Mac: ⌘ + .
Windows/Linux: Ctrl + .

/qa: For executing test cases post-review. /ask: To explain review results or clarify issues. /coach: For converting reviews into learning goals.

Author Notes​

This mode was created to bridge engineering accuracy with real-world requirements, combining deep static analysis, contextual interpretation, and compliance scanning into one seamless review workflow — all powered by LLMs and framework-aware intelligence. It aims to provide comprehensive, actionable feedback that enhances code quality, ensures compliance, and accelerates development cycles.