← Claude Skills Browser

Claude Code /review Command

How to run automated code review with Claude Code — using /review, /ultrareview, and /security-review.

/review

Deep PR Analysis — Built into Claude Code

Reads all files changed on the current branch vs. main. Returns a structured review: logic bugs, security flags, style issues, missed edge cases, and API misuse — with severity levels and concrete fix suggestions.

Three review commands: which to use when

Command Best for How it works Notes
/review Standard PR review Single model, current branch vs. main Free, fastest
/ultrareview Critical merges, large PRs Multiple agents in parallel — security, logic, style, tests each get dedicated analysis Billed separately; use /ultrareview 123 for GitHub PR #123
/security-review Auth, payments, user input OWASP Top 10 focused — injection, auth bypass, unsafe deps, insecure defaults Free; run before any deploy touching user data

How to use /review

Type /review at the start of a message in any Claude Code session while on a feature branch. Claude Code will automatically diff against your main branch and analyze all changed files.

Example — standard review
/review
Example — review a specific GitHub PR
/ultrareview 42
Example — loop review as you code
/loop 10m /review

What /review checks

Logic bugs

Off-by-one errors, null dereferences, incorrect loop termination, wrong operator precedence, control-flow issues that only surface in edge cases.

Security issues

SQL/command injection vectors, missing input sanitization, insecure direct object references, hardcoded credentials, missing auth checks on routes.

API misuse

Incorrect use of external library APIs, missing required cleanup (file handles, DB connections), wrong concurrency patterns (race conditions, deadlocks in async code).

Style & maintainability

Inconsistent naming, overly complex functions that should be extracted, dead code, missing tests for new logic.

Automating code review with /loop and /schedule

Use /loop to run /review continuously as you code. Combine with /schedule to set up daily automated branch reviews.

Continuous review while coding
/loop 15m /review
Schedule nightly review
/schedule run /security-review on this branch every night at 11pm

Browse All Claude Code Commands

See /review, /security-review, /ultrareview, /loop and 18 more skills — searchable, with examples.

Open Claude Skills Browser →

FAQ

What does the /review command do in Claude Code?
The /review command triggers a deep multi-file pull request review. It diffs the current branch against main, reads all changed files, and surfaces logic bugs, security issues, style concerns, API misuse, and missed edge cases. Output is structured by severity with suggested fixes.
What is /ultrareview in Claude Code?
/ultrareview launches a multi-agent cloud review of the current branch or a specific GitHub PR. Multiple specialized agents work in parallel, each examining a different concern (security, logic, style, tests). Use /ultrareview for the current branch, or /ultrareview 123 to review GitHub PR #123. It is billed separately from regular Claude usage.
When should I use /review vs /ultrareview vs /security-review?
Use /review for standard PR review — quick single-model analysis. Use /ultrareview when you need deeper parallel coverage for critical merges or large PRs. Use /security-review when changes touch auth, payments, user data, or input parsing — it focuses exclusively on OWASP Top 10 and security-critical patterns.
Can I automate /review to run on every commit?
Yes — use /loop to run /review on a recurring interval while you're coding: /loop 10m /review runs a review every 10 minutes. For CI-style automation, /schedule can run /review on a cron schedule. Use /update-config to add hooks that trigger /review on specific events.

← All slash commands   |   Automating tasks with /loop & /schedule →

More free AI dev tools

⚡ Using Claude Code? 30 power prompts that 2× your output · £5 £3 first 10Get PDF £3 →