external-code-review

Reviews a diff via external LLM returning structured JSON verdicts and concerns with suppression support.

2|1|Updated Apr 9, 2026
One-click install
npx skills add https://github.com/timgranlundmarsden/claude-agent-flow --skill external-code-review
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: external-code-review
Source: https://github.com/timgranlundmarsden/claude-agent-flow/tree/main/skills/external-code-review
Command: npx skills add https://github.com/timgranlundmarsden/claude-agent-flow --skill external-code-review

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires curl, git, jq, python3, pyyaml, and includes scripts (resource) components.

What problem does it solve?

This Skill automates thorough code review by sending a diff to an external LLM and returning a structured verdict, summary, and per-file concerns. It helps teams catch blocking issues, warnings, and review notes without hand-parsing model output.

Core Features & Use Cases

  • External review execution: Sends a diff to the configured API endpoint and requests JSON-only review output.
  • Robust response handling: Validates required environment variables, retries transient HTTP 503 failures, and parses messy model responses into clean JSON.
  • Suppression support: Applies file-and-keyword suppression rules so known issues can be acknowledged without blocking the review.
  • Use case: Run this skill on a pull request diff to get an automated second-pass review before merge.

Quick Start

Use the external-code-review skill to review this diff and return a structured JSON verdict with any concerns.

Frequently Asked Questions about external-code-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate code review for a pull request diff using an LLM?

Automate code review for a pull request diff by sending the diff to an external LLM API endpoint, which returns a structured JSON verdict with blocking, warning, and informational concerns.

Can I suppress known issues in an automated LLM code review?

You can suppress known issues during an LLM code review by applying file-and-keyword suppression rules, allowing acknowledged issues to be ignored when recalculating the final verdict.

What environment variables are needed for external LLM API integration in code review?

External LLM API integration for code review requires specific environment variables for API configuration, which the script validates before executing the review to ensure a successful connection.

How does the diff analysis handle transient API failures during code review?

Diff analysis handles transient API failures during code review by automatically retrying HTTP 503 responses, ensuring the external LLM request completes despite temporary server unavailability.

Does the code review skill parse messy LLM model output into JSON?

The code review skill parses messy model responses from the LLM into clean JSON-only output, extracting structured findings and per-file concerns without requiring manual parsing.

Do I need curl and jq to run an external LLM code review on my repository?

You need curl, jq, git, python3, and pyyaml installed to run the external LLM code review, as these dependencies support diff extraction, API requests, and JSON response parsing.