README

Security Audit Skill

Security audit patterns (OWASP Top 10, CWE Top 25 2025, CVSS v4.0) and GitHub project security checks for any project. Deep automated PHP/TYPO3 code scanning with 80+ checkpoints, 41 reference guides, and PreToolUse warnings.

Compatibility

This is an Agent Skill following the open standard originally developed by Anthropic and released for cross-platform use.

Supported Platforms:

Claude Code (Anthropic)
Cursor
GitHub Copilot
Other skills-compatible AI agents

Skills are portable packages of procedural knowledge that work across any AI agent supporting the Agent Skills specification.

Features

Vulnerability Assessment: XXE injection, SQL injection, XSS, CSRF, command injection, path traversal, file upload vulnerabilities, insecure deserialization, SSRF, type juggling, SSTI, JWT flaws, LDAP injection, email header injection, session fixation
Risk Scoring: CVSS v3.1 and v4.0 scoring methodology, risk matrix assessment, impact and likelihood analysis, prioritization frameworks
Secure Coding: Input validation, output encoding, cryptographic best practices (sodium), session management, authentication patterns, security headers
Standards Compliance: OWASP Top 10, CWE Top 25 (2025), OWASP ASVS v4.0, Proactive Controls — applicable to any project
PHP/TYPO3 Deep Scanning: 80+ automated checkpoints, PHP 8.x security features, framework patterns (TYPO3, Symfony, Laravel)
DevSecOps: CI/CD security pipeline, SAST, dependency scanning, supply chain security, SLSA

Installation

Marketplace (Recommended)

Add the Netresearch marketplace once, then browse and install skills:

# Claude Code
/plugin marketplace add netresearch/claude-code-marketplace
/plugin install security-audit@netresearch-claude-code-marketplace

Without a marketplace

Since Claude Code 2.1.157 a plugin directory under your personal skills directory loads on its own, including the hooks this repo ships:

mkdir -p ~/.claude/skills
git clone https://github.com/netresearch/security-audit-skill.git \
  ~/.claude/skills/security-audit

It loads as security-audit@skills-dir on the next session. Update with git -C ~/.claude/skills/security-audit pull and start a new session; remove it by deleting the directory. This route has no claude plugin update.

npx (skills.sh)

Install with any Agent Skills-compatible agent:

npx skills add https://github.com/netresearch/security-audit-skill --skill security-audit

Limitation: npx skills installs SKILL.md-based skills only. This repo also ships hooks, which it does not install — use the marketplace or the skills directory for those.

Download Release

Download the latest release and extract to your agent's skills directory.

Git Clone

git clone https://github.com/netresearch/security-audit-skill.git

Composer (PHP Projects)

composer require netresearch/security-audit-skill

Requires netresearch/composer-agent-skill-plugin.

npm (Node Projects)

npm install --save-dev \
  @netresearch/agent-skill-coordinator \
  github:netresearch/security-audit-skill

Requires @netresearch/agent-skill-coordinator, which discovers the skill in node_modules and registers it in AGENTS.md via a postinstall hook. For pnpm, also allowlist the coordinator's postinstall:

{
  "pnpm": {
    "onlyBuiltDependencies": ["@netresearch/agent-skill-coordinator"]
  }
}

Usage

This skill is automatically triggered when:

Conducting security assessments
Identifying vulnerabilities (XXE, SQL injection, XSS, CSRF, command injection)
Scoring security risks with CVSS v3.1 or v4.0
Implementing secure coding practices
Auditing PHP applications for security issues
Reviewing code for OWASP Top 10 vulnerabilities
Setting up CI/CD security pipelines

Example queries:

"Audit this code for XXE vulnerabilities"
"Check for SQL injection risks"
"Score this vulnerability using CVSS v4.0"
"Review authentication implementation for security flaws"
"Implement secure XML parsing"
"What security headers should this application set?"

Structure

security-audit-skill/
├── SECURITY.md                           # Security policy
├── hooks/
│   └── hooks.json                        # PreToolUse hook configuration
├── scripts/
│   ├── check_risky_command.py            # Risky command detection hook
│   ├── validate_checkpoints.py           # checkpoints.yaml structure check
│   └── test_*.py                         # Tests (run by ci.yml)
├── skills/security-audit/
│   ├── SKILL.md                          # Skill definition
│   ├── checkpoints.yaml                  # 80+ automated security checkpoints
│   ├── scripts/
│   │   ├── security-audit-dispatcher.sh  # Detects the stack, runs the matching scanners
│   │   ├── scanners/                     # Per-ecosystem scanner modules
│   │   ├── security-audit.sh             # PHP project security audit
│   │   └── github-security-audit.sh      # GitHub repo security audit
│   └── references/                       # 41 reference guides
└── .github/
    └── workflows/
        ├── ci.yml                        # Python tests
        ├── lint.yml                      # Skill validation, linters, ShellCheck
        ├── security.yml                  # Secret, workflow, dependency and SAST scans
        └── release.yml                   # Release automation

Expertise Areas

Vulnerability Assessment

XXE (XML External Entity) injection detection
SQL injection pattern recognition
XSS (Cross-Site Scripting) analysis
CSRF protection verification
Command injection detection
Path traversal prevention
File upload security
Insecure deserialization
SSRF detection
Authentication/authorization flaws

Risk Scoring

CVSS v3.1 scoring methodology
CVSS v4.0 scoring methodology
Risk matrix assessment
Impact and likelihood analysis
Prioritization frameworks

Secure Coding

Input validation patterns
Output encoding strategies
Secure configuration
Cryptographic best practices (sodium)
Session management
Authentication patterns (Argon2, JWT, MFA)
Security headers (HSTS, CSP)

DevSecOps

SAST integration (PHPStan, Semgrep, CodeQL)
Dependency scanning (composer audit, Trivy)
Supply chain security (SLSA, Sigstore)
Container security (Hadolint, Trivy)
SBOM generation (CycloneDX)

Security Audit Checklist

Authentication & Authorization

Password hashing uses bcrypt/Argon2 (PASSWORD_ARGON2ID)
Session tokens are cryptographically random (random_bytes)
Session fixation protection enabled (session_regenerate_id)
CSRF tokens on all state-changing operations
Authorization checks on all protected resources
Rate limiting on authentication endpoints

Input Handling

All input validated server-side
Parameterized queries for all SQL
XML parsing with external entities disabled (LIBXML_NONET only)
File uploads restricted by type (MIME validation) and size
Path traversal prevention on file operations
No unserialize() with user input

Output Handling

Context-appropriate output encoding (htmlspecialchars)
Content-Type headers set correctly
X-Content-Type-Options: nosniff
Content-Security-Policy configured
X-Frame-Options or CSP frame-ancestors set
Strict-Transport-Security (HSTS) enabled

Data Protection

Sensitive data encrypted at rest (sodium_crypto_secretbox)
TLS 1.2+ for data in transit
Secrets not in version control
PII handling compliant with regulations
Audit logging for sensitive operations

Related Skills

enterprise-readiness-skill: References this skill for security assessment
php-modernization-skill: Type safety enhances security
typo3-testing-skill: Security test patterns

Development

Run the tests from the repository root with uv:

uv run scripts/test_risky_patterns.py      # hook patterns: expected matches and non-matches
uv run scripts/test_check_risky_command.py # hook input and output as Claude Code uses them
uv run scripts/test_audit_scripts.py       # audit scripts against fixtures (needs bash 4+, GNU grep, jq)
uv run scripts/validate_checkpoints.py     # checkpoints.yaml structure and unique IDs
pre-commit run --all-files                 # the linters CI runs in Skill Validation

CI runs the three test scripts on Python 3.12 and 3.13 (.github/workflows/ci.yml). A failing test prints its name and the assertion that failed; the test scripts exit non-zero. A change to the hook or the audit scripts comes with a test in these files.

Dependencies

Using the skill: the audit scripts need bash 4+, GNU grep (for grep -P) and find; github-security-audit.sh needs the gh CLI. The hook needs python3 and uses only its standard library. composer, npm, govulncheck and trufflehog are optional: a scanner calls them only when command -v finds them, and the skill does not install them.
Installing the skill: composer.json requires netresearch/composer-agent-skill-plugin (^2.0); package.json declares @netresearch/agent-skill-coordinator as a peer dependency. The repository commits no lock file for either; the consuming project resolves them.
Development and CI: scripts/validate_checkpoints.py declares PyYAML as inline script metadata, which uv run installs; the test scripts use the standard library. The pre-commit hooks are pinned by rev: in .pre-commit-config.yaml. The workflows call reusable workflows from netresearch/skill-repo-skill, netresearch/.github and netresearch/typo3-ci-workflows at @main; the third-party actions inside those are pinned to commit SHAs there.
Tracking: Renovate (renovate.json, config:recommended with pre-commit updates enabled) opens update pull requests, and .github/workflows/auto-merge-deps.yml hands them to the organisation's auto-merge workflow. On pull requests, dependency review and Composer Audit check the dependencies (.github/workflows/security.yml). New dependencies must meet the licence and vulnerability rules of the organisation's security policy linked below.

Governance and policies

This repository follows the Netresearch organisation policies:

Governance: ownership, roles, how decisions are made and disputes resolved, and continuity.
Roadmap: planned and explicitly excluded work for the coming year.
Handling of dependency and code analysis findings: thresholds, deadlines and the exception process for dependency (SCA) and static analysis (SAST) findings.
Secret management: how CI and release credentials are stored, accessed and rotated.
Access roster: who holds administrative access to this repository and the organisation.

The architecture of this skill (actors, components, data flows) is described in docs/ARCHITECTURE.md, and its security assurance case (threat model, trust boundaries, countermeasures and limits) in docs/SECURITY-ASSURANCE.md. Supported versions and vulnerability reporting: SECURITY.md.

Checks that run on every pull request to main in this repository:

Security (.github/workflows/security.yml): Betterleaks secret scanning, zizmor workflow analysis, dependency review, and, through the typo3-ci-workflows security reusable, Composer Audit and Opengrep SAST (findings handled as the organisation's static analysis rule sets out).
Skill Validation (.github/workflows/lint.yml): skill structure, plugin manifest sync, markdownlint, yamllint, actionlint, JSON syntax, version checks, ShellCheck at the validator's default severity error, ruff and the checkpoint schema. The pre-commit hook runs ShellCheck at style.
CI (.github/workflows/ci.yml): the tests listed under Development.
Eval Validation, Harness Verification, the template drift check, Labeler and PR Quality Gates.
CodeQL through GitHub's default setup, a repository setting.
Configured outside the workflows: DCO sign-off check, SonarCloud code analysis, CodeRabbit review, and GitHub secret scanning with push protection, which rejects pushes that contain a recognised secret.

Which of these checks must pass before a merge is set in the repository's branch protection, not in this repository.

License

This project uses split licensing:

Code (scripts, workflows, configs): MIT
Content (skill definitions, documentation, references): CC-BY-SA-4.0

See the individual license files for full terms.

Credits

Developed and maintained by Netresearch DTT GmbH.

Made with love for Open Source by Netresearch

netresearch / agent-security-audit

Package info

Security

Statistics