Security Audit

Security Audit

Investigate the codebase for security vulnerabilities across OWASP Top 10 categories, secrets exposure, supply chain risks, and language-specific patterns. Produces a scored SECURITY_AUDIT.md with prioritized remediation actions.

Core principle: Investigate, do not just grep. Read surrounding code to determine if a match is a genuine vulnerability, a false positive, or already mitigated. Only flag findings with specific file:line evidence and confirmed risk.

Prerequisite check

git --version 2>/dev/null && git rev-parse --is-inside-work-tree 2>/dev/null

Check for available security tools (optional but enhance analysis):

command -v gitleaks 2>/dev/null && echo "gitleaks: available"
command -v semgrep 2>/dev/null && echo "semgrep: available"
command -v trivy 2>/dev/null && echo "trivy: available"
command -v govulncheck 2>/dev/null && echo "govulncheck: available"

Proceed without tools — manual investigation still works. Note which tools are missing in the report.

Step 1: Scope and context

Determine what to audit and understand the application:

1. Detect tech stack — scan for go.mod, package.json,

pyproject.toml, Cargo.toml, *.csproj to identify languages

1. Identify security-sensitive areas — search for files

related to authentication, authorization, database access, API endpoints, cryptography, file handling

1. Determine scope — full codebase audit, specific module, or

recent changes only (based on user input)

1. Read project docs — CLAUDE.md, README.md, CONTEXT.md for

architectural context and known boundaries

Step 2: Automated scanning

Run available security tools before manual investigation. Tool findings guide where to focus manual review.

# Secrets in current code and git history
gitleaks detect -v 2>&1

# Static analysis (language-aware rule sets)
semgrep --config=auto --severity=ERROR . 2>&1

# Dependency vulnerabilities
trivy fs --severity HIGH,CRITICAL . 2>&1
# or: govulncheck ./... / npm audit / pip-audit / cargo audit

If tools find issues, verify each one — automated scanners produce false positives. Read the flagged code in context before including in the report.

Step 3: OWASP Top 10 analysis

Systematically investigate each OWASP category relevant to the project. Not all categories apply to all projects — a CLI tool does not need XSS checks; a library does not need CSRF protection.

For each applicable category:

1. Identify the code paths where this vulnerability could exist
2. Read the actual implementation (not just grep for keywords)
3. Check if mitigations are in place
4. Flag only confirmed or strongly suspected vulnerabilities

For detailed investigation patterns per OWASP category, see references/owasp-patterns.md.

Step 4: Secrets and supply chain

Secrets: Scan current code AND git history for hardcoded credentials, API keys, tokens, and connection strings. A secret removed from current code but present in git history is still compromised — flag as CRITICAL.

Supply chain: Check lockfile integrity, dependency vulnerabilities, dependency confusion risk, and maintenance signals for critical dependencies.

For detailed scanning commands and analysis patterns, see references/secrets-and-supply-chain.md.

Step 5: Threat modeling

Apply STRIDE to the actual codebase:

1. Map entry points — every way data enters the application
2. Identify trust boundaries — where untrusted data crosses

into trusted zones

1. Apply STRIDE categories — Spoofing, Tampering, Repudiation,

Information Disclosure, Denial of Service, Elevation of Privilege

1. Score threats using DREAD (Damage, Reproducibility,

Exploitability, Affected users, Discoverability)

For detailed threat modeling steps and DREAD scoring, see references/threat-model.md.

Step 6: Language-specific review

Check for vulnerability patterns specific to the detected languages. Load only the relevant language sections.

For Go, JavaScript/TypeScript, Python, Rust, and .NET patterns, see references/language-security.md.

Step 7: Score and classify findings

For each finding, assign:

• Severity: CRITICAL / HIGH / MEDIUM / LOW
• CRITICAL: exploitable now, data breach or RCE risk
• HIGH: exploitable with moderate effort, significant impact
• MEDIUM: requires specific conditions, moderate impact
• LOW: theoretical risk, minimal impact
• Effort to fix: S (< 1 day) / M (1-5 days) / L (1-2 weeks)
• Evidence: specific file:line reference and explanation

Step 8: Generate SECURITY_AUDIT.md

Write the report using the template in references/report-template.md. Include executive summary, detailed findings with evidence, threat model summary, and prioritized remediation actions.

Run markdownlint on the generated file.

Step 9: Present summary

Show the user:

1. Finding counts by severity
2. Top 3 most critical items with evidence
3. Tools that were available vs missing
4. Remediation priority list
5. Ask which findings to address first

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: rshade
• Source: rshade/agent-skills
• License: Apache-2.0

Install and usage instructions live in the source repository linked above.