AI-Powered Code Review

VerifiedSafe

Exhaustive code analysis using OpenAI Codex with GitHub/Linear/Sentry issue context. Perfect before creating PRs or after completing features.

Sby Skills Guide Bot
DevelopmentIntermediate
406/2/2026
Claude Code
#code-review#ai-codex#development

Recommended for

Our review

Performs an exhaustive AI-powered code review using OpenAI Codex, gathering context from GitHub, Linear, and Sentry issues.

Strengths

  • Deep analysis with external context integration (issues, plans)
  • Supports multiple modes: uncommitted, branch, specific commit
  • Can gather context without running the full review

Limitations

  • Requires the `review.py` script and Codex CLI to be installed and authenticated
  • Depends on external APIs (GitHub, Linear, Sentry) via access tokens
  • Only works for projects with a Git repository
When to use it

Use this skill to get an automated, context-aware code review before creating a pull request or after completing feature work.

When not to use it

Do not use it for trivial changes or non-code tasks (e.g., pure documentation).

Security analysis

Safe
Quality score88/100

The skill runs a local Python script for AI-powered code review, which reads environment variables for legitimate API access and does not execute destructive or obfuscated commands. No curl piping to shell, no exfiltration, no disabling of safety features.

No concerns found

Examples

Review uncommitted changes
Review my changes
Review with issue context
/review --issues "#123"
Review against main branch
review my changes against main

name: review description: | Exhaustive code review using OpenAI Codex. Use when user says "review", "/review", "review my changes", or wants AI-powered code analysis. Gathers context from GitHub/Linear/Sentry issues. allowed-tools: Bash, Read, Glob, Grep

Review

Exhaustive AI-powered code review via OpenAI Codex.

When to Use

  • User says "review", "/review", "review my changes"
  • Before creating PR
  • After completing feature work

Workflow

1. Gather Issue References

Scan conversation for:

  • GitHub: #123, org/repo#123, GitHub URLs
  • Linear: PROJ-123, Linear URLs
  • Sentry: sentry:ID, Sentry URLs

2. Read Plan File (if exists)

Check for active plan in conversation or ~/.claude/plans/

3. Read Referenced Files

If user mentions @file.ts or specific files, include content.

4. Run Review

# Default: reviews uncommitted changes if any exist
./scripts/review.py codex \
  --issues "#123,PROJ-456" \
  --plan ./plan.md \
  --files "src/api.ts"

# If no uncommitted changes, prompts for branch/commit to compare

Command

./scripts/review.py codex [OPTIONS]

Options

| Option | Description | |--------|-------------| | --base, -b BRANCH | Compare against branch | | --uncommitted, -u | Review staged/unstaged/untracked changes (default if changes exist) | | --commit, -c SHA | Review specific commit | | --issues, -i REFS | Issue refs: #123, PROJ-456, sentry:ID, or URLs | | --plan, -p PATH | Plan file for context | | --files, -f PATHS | Additional files (comma-separated) | | --title, -t TEXT | Commit/PR title for summary | | --model, -m MODEL | Codex model (default: gpt-5.1-codex-max) |

Context Command

Gather context without running review:

./scripts/review.py context [OPTIONS]

| Option | Description | |--------|-------------| | --issues, -i REFS | Issue refs (comma-separated) | | --plan, -p PATH | Plan file path | | --files, -f PATHS | Additional files (comma-separated) | | --output, -o FORMAT | json or markdown (default: markdown) |

Requirements

  • codex CLI installed and authenticated
  • LINEAR_API_KEY - (optional) Linear API key for Linear issues
  • SENTRY_AUTH_TOKEN - (optional) Sentry auth token
  • SENTRY_ORG - (optional) Sentry org slug

Output

The script displays gathered context to stderr, then launches codex review interactively. Codex outputs its review directly to the terminal.

Examples

# Default: review uncommitted changes (or prompt if none)
./scripts/review.py codex

# Explicitly review uncommitted changes
./scripts/review.py codex --uncommitted

# Review changes against specific branch
./scripts/review.py codex --base main

# Review with GitHub issue context
./scripts/review.py codex --issues "#123"

# Review with plan file
./scripts/review.py codex --plan ./plan.md

# Gather context as JSON (without running review)
./scripts/review.py context --issues "#123" --output json
Related skills