cross-review

Runs an advisory cross-vendor code review of the current branch diff against main.

1|1|Updated Feb 2, 2026
One-click install
npx skills add https://github.com/corticalstack/flow --skill cross-review-corticalstack
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: cross-review
Source: https://github.com/corticalstack/flow/tree/main/.claude/skills/cross-review
Command: npx skills add https://github.com/corticalstack/flow --skill cross-review-corticalstack

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Single-model code reviews can miss bugs that a model from a different vendor lineage would catch. This Skill sends your branch's diff vs main to a configurable second-opinion backend (Azure AI Foundry, GitHub Models, or Copilot CLI) and returns advisory HIGH/CRITICAL findings without modifying any code. ## Core Features & Use Cases - Multi-backend reviewer profiles: Choose between Azure Foundry OpenAI endpoints, legacy Azure AI Inference, GitHub Models, or the Copilot CLI, all authenticated via OAuth with no static API keys. - Advisory-only findings: The reviewer reports only HIGH or CRITICAL issues (security, correctness, data loss) and never applies changes; results are saved as markdown under flow/reviews/. - Use Case: Before merging a feature branch, run /cross-review with the github-o3-mini profile to get a second opinion on the diff, then decide yourself which findings warrant fixes. ## Quick Start Ask the assistant to run a cross-vendor review of the current branch using the default profile, after copying profiles.example.json to profiles.json and completing the required OAuth login.

Frequently Asked Questions about cross-review

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

FAQPage Schema
How do I run a cross-vendor code review on my branch?▼

Invoke /cross-review optionally with a profile name. The skill computes the diff vs main, sends it to the configured reviewer backend, prints HIGH/CRITICAL findings, and saves a markdown copy under flow/reviews/.

Which AI backends can act as the second-opinion reviewer?▼

Four backends are supported: Azure AI Foundry OpenAI v1 endpoints, legacy Azure AI Inference (deprecated), GitHub Models, and the GitHub Copilot CLI. All use OAuth tokens from az, gh, or copilot logins, so no static API keys are stored.

Does the cross-review skill modify my code automatically?▼

No. The review is strictly advisory: findings are surfaced to the user and saved to a file, but no code changes are applied. The Copilot CLI backend is even invoked with write tools denied to keep the review read-only.

Why does cross-review fail with an authentication error?▼

Each backend needs its own OAuth login: run az login for Azure Foundry, gh auth login for GitHub Models, or copilot auth login for Copilot CLI. The script also requires jq, curl, and git to be installed.

What happens if my branch has no diff against main?▼

The script detects the empty diff and exits with an informational message that there is nothing to review. You can set CROSS_REVIEW_BASE to a different base ref if your changes are relative to another branch.