Claude Code Skill · $19 one-time

5 passes. 0 bugs merged.

The code review that catches what you miss. A structured 5-pass review protocol for Claude Code — correctness, security, performance, maintainability, and test coverage.

Your PR reviews are missing critical bugs

You ask Claude Code to “review my PR.” It says “looks clean, maybe add some tests.” You merge. Something breaks at 2am.

Without Deep PR Review
# "Review my PR"

✓ Code looks clean
✓ Consider adding some tests
✓ Minor: rename `data` to something descriptive

# Meanwhile, in production...
💥 SQL injection in search endpoint
💥 XSS via user-supplied HTML
💥 N+1 query taking down the DB
With Deep PR Review
## PR Review: Add search endpoint
Verdict: REQUEST_CHANGES

[SECURITY] critical — search.ts:28
  User input interpolated into SQL
  Fix: Use parameterized query

[SECURITY] critical — render.tsx:15
  dangerouslySetInnerHTML on user data
  Fix: Sanitize with DOMPurify

[PERFORMANCE] high — search.ts:42
  N+1 query in results loop
  Fix: Batch with IN clause

5 focused passes. Every PR.

Senior engineers don't review code in a single sweep. They check correctness, then security, then performance — each as a focused pass. Deep PR Review does the same.

Pass 1

Correctness

Logic errors, edge cases, race conditions, type safety

Pass 2

Security

SQL injection, XSS, missing auth, data exposure (OWASP)

Pass 3

Performance

N+1 queries, missing indexes, unbounded ops, memory leaks

Pass 4

Maintainability

Naming, complexity, dead code, codebase consistency

Pass 5

Test Coverage

Specific missing test scenarios — not generic 'add tests'

Real bugs caught

These passed a quick “review this” prompt. They wouldn't pass Deep PR Review.

SECURITYcritical

SQL injection — in a "simple" 45-line search endpoint

SECURITYcritical

XSS — via dangerouslySetInnerHTML on user-supplied data

SECURITYhigh

IDOR — letting any user view another user's analytics

CORRECTNESShigh

Stale closure — from a missing useEffect dependency

PERFORMANCEhigh

Full table scan — on every search request (leading wildcard LIKE)

Install in 30 seconds

One file. Drop it in. Every PR gets a senior-level review.

terminal
# After purchase, unzip and copy
mkdir -p .claude/skills
cp deep-pr-review.md .claude/skills/

# That's it. Now ask Claude:
"Review my PR"

Who this is for

Solo devs

A second pair of eyes before every merge. No team required.

Team leads

Consistent review quality across the team. No more “LGTM” rubber stamps.

Startups

Ship fast without a dedicated security reviewer. Every PR gets OWASP-aligned checks.

What the output looks like

Structured, severity-rated, actionable. Every finding has a file, line number, and fix.

deep-pr-review output
## PR Review: Add user authentication middleware

Verdict: REQUEST_CHANGES

Summary: Adds JWT-based auth middleware.
Core implementation is solid, but critical
security issue with token validation.

### Critical Issues
[SECURITY] critical — auth.ts:28
  JWT secret fallback to "secret" string.
  In production, all tokens signed with
  known value if env var missing.
  Fix: Remove fallback. Throw on startup.

### Recommendations
[CORRECTNESS] high — auth.ts:45
  Token expiry uses < instead of <=
  Fix: Use jwt.verify built-in expiry

[TESTING] high
  No test for expired token rejection.
  Suggested: assert 401 for exp = now - 1

One file. Five passes. Zero bugs merged.

Stop merging bugs. Start shipping confidence. $19 one-time — works with any language, any project.

Instant download · Works with Claude Code · MIT License