review

Reviews pull request diffs for SQL safety, race conditions, and LLM trust boundary violations.

Updated Jun 22, 2026
One-click install
npx skills add https://github.com/aicodepro/ai-agent-nexi --skill review-aicodepro
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: review
Source: https://github.com/aicodepro/ai-agent-nexi/tree/main/agent/skills/gstack/review
Command: npx skills add https://github.com/aicodepro/ai-agent-nexi --skill review-aicodepro

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Catching structural defects like SQL injection, race conditions, shell injection, and unvalidated LLM output before code lands is time-consuming and error-prone when done manually. This Skill automates pre-landing PR review by analyzing the diff against the base branch with a structured, severity-tiered checklist. ## Core Features & Use Cases - Two-Pass Diff Review: Runs a CRITICAL pass (SQL & data safety, race conditions, LLM output trust boundary, shell injection, enum completeness) followed by an INFORMATIONAL pass (async/sync mixing, type coercion, CI/CD pipeline issues, and more). - Parallel Specialist Subagents: Dispatches dedicated checklists for testing gaps, maintainability, security, performance, data migrations, API contracts, and an adversarial red-team pass on large diffs. - Fix-First Workflow: Auto-applies mechanical fixes (dead code, N+1 queries, magic numbers) while batching genuinely ambiguous findings into a single structured user decision. - Greptile Comment Triage: Fetches, classifies, and replies to Greptile bot review comments on GitHub PRs with evidence-backed responses. - Use Case: Before merging a feature branch, ask for a pre-landing review; the Skill inspects git diff origin/main, flags a missing eager-load causing N+1 queries, auto-fixes it, and asks you to confirm a race-condition fix. ## Quick Start Ask the assistant to review this PR or run a pre-landing review of my current diff against the base branch.

Frequently Asked Questions about review

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

FAQPage Schema
How do I run a pre-landing code review on my PR diff?

Invoke the review skill with a request like "review this PR" or "check my diff". It runs git diff against the base branch, applies a two-pass severity-tiered checklist, auto-fixes mechanical issues, and asks you about ambiguous findings.

What issues does an automated diff review check for?

The critical pass checks SQL injection, race conditions, LLM output trust boundaries, shell injection, and enum completeness. The informational pass covers async/sync mixing, column name safety, type coercion, time windows, and CI/CD pipeline problems.

Can automated review triage Greptile bot comments on GitHub PRs?

Yes. The skill fetches Greptile line-level and top-level comments via the GitHub CLI, classifies each as valid, already fixed, false positive, or suppressed, and posts evidence-backed replies using tiered templates.

Does the review skill auto-fix findings or always ask first?

It uses a fix-first heuristic: mechanical fixes like dead code removal, N+1 eager loading, and magic numbers are applied automatically, while security issues, race conditions, and behavior changes are batched into a single user question.

When does the design review checklist run during a diff review?

The design checklist runs only when the diff touches frontend files, detected via the gstack-diff-scope tool. If no frontend files changed, the design review is skipped silently with no output.