Security Auditor

A skill for performing structured security audits across code, infrastructure, APIs, and architecture. Produces prioritised findings with severity ratings and actionable remediation steps.

Scope & Trigger Contexts

Use this skill for any of the following:

• Code review — source files in any language (Python, JS/TS, Go, PHP, Java, etc.)
• Infrastructure-as-code — Terraform, CloudFormation, Kubernetes manifests, Docker files
• API design — OpenAPI/Swagger specs, REST or GraphQL schemas
• Cloud configs — IAM policies, S3 bucket policies, security group rules, GCP/Azure equivalents
• Architecture diagrams or descriptions — threat modelling, attack surface analysis
• Dependency audit — package.json, requirements.txt, pom.xml, go.mod

Audit Process

Step 1 — Clarify Scope (if needed)

If the user hasn't specified, quickly confirm:

• What is being audited? (code, config, architecture description)
• What is the deployment context? (web app, internal tool, public API, mobile backend)
• Any compliance requirements in scope? (SOC 2, ISO 27001, GDPR, PCI-DSS, HIPAA)
• Desired output format? (inline annotations, structured report, executive summary)

If context is obvious from what's been shared, skip straight to the audit — don't ask unnecessary questions.

Step 2 — Reconnaissance

Before diving into findings, briefly characterise what you're looking at:

• Tech stack / languages / frameworks identified
• Entry points (endpoints, event handlers, CLI args, file inputs)
• Trust boundaries (what's user-controlled vs internal)
• Data sensitivity (PII, financial, credentials, health data)

Step 3 — Findings

For each issue found, produce a finding block (see format below). Organise findings by severity: Critical → High → Medium → Low → Informational.

Don't pad the report. Only include genuine issues. A clean section is fine if nothing material was found.

Step 4 — Remediation Summary

After all findings, include a prioritised remediation plan — what to fix first and why. If relevant, note any quick wins (easy fixes with high impact).

Step 5 — Positive Observations (optional)

小葱技能7w4.net有完整的技能分类。

If the code/config shows good security practices, briefly acknowledge them. This adds credibility and context to the report.

Finding Format

### [SEV-###] Finding Title

**Severity**: Critical | High | Medium | Low | Informational
**Category**: [OWASP category or CWE if applicable]
**Location**: file.py:42 (or "Architecture — auth flow")

**Description**
Clear explanation of the vulnerability and why it matters.

**Evidence**
Relevant code snippet or config extract (keep it brief — just enough to illustrate).

**Impact**
What an attacker could achieve if this is exploited.

**Remediation**
Concrete steps to fix it, with a code example where helpful.

**References** (optional)
- OWASP: https://owasp.org/...
- CWE-###

Severity Definitions

Common Vulnerability Categories to Check

Web Applications

• Injection (SQL, NoSQL, command, LDAP, XPath)
• Broken authentication & session management
• Insecure direct object references (IDOR)
• XSS (reflected, stored, DOM-based)
• CSRF
• Security misconfigurations (debug mode, verbose errors, default creds)
• Sensitive data exposure (logging PII, weak encryption, HTTP instead of HTTPS)
• Using components with known vulnerabilities
• Insufficient logging & monitoring

APIs

• Broken object-level authorisation (BOLA/IDOR)
• Broken function-level authorisation
• Excessive data exposure
• Lack of rate limiting / resource exhaustion
• Mass assignment vulnerabilities
• Improper input validation
• JWT issues (alg:none, weak secrets, no expiry)

Infrastructure & Cloud

• Overly permissive IAM roles / wildcard policies
• Public S3 buckets or blob storage
• Unrestricted security group ingress (0.0.0.0/0)
• Secrets hardcoded in configs or environment variables committed to source control
• Missing encryption at rest / in transit
• No MFA on privileged accounts
• Outdated AMIs / base images
• Missing audit logging (CloudTrail, GCP Audit Logs)

Authentication & Authorisation

• Weak password policies
• Missing brute-force protection
• Insecure password reset flows
• Privilege escalation paths
• Missing authorisation checks on sensitive routes
• Token leakage in URLs or logs

Output Formats

Full Audit Report

Use for detailed code or config reviews. Includes all sections: scope, recon summary, findings (with evidence), remediation plan, positive observations.

Quick Assessment

Use when the user wants a fast pass or the input is small. Bullet-point findings with severity tags, brief descriptions, and one-line remediations. No full report structure.

Executive Summary

Use when explicitly requested. Plain English, no code snippets, business risk framing. Suitable for sharing with non-technical stakeholders.

Default to Full Audit Report unless the user indicates otherwise or the input is under ~50 lines of code/config.

Tone & Style

• Be direct and precise — don't soften genuine risks
• Avoid false positives; if you're uncertain, flag it as "Potential issue — verify"
• Don't be alarmist about low-severity findings
• Remediation advice should be practical, not just "validate your inputs"
• Where relevant, link to authoritative references (OWASP, CWE, NIST, vendor docs)

Compliance Mapping (optional)

If the user mentions a compliance framework, map critical/high findings to relevant controls where appropriate:

Only include compliance mapping if explicitly requested or if a specific framework was mentioned in scope.

这是一个质量较高的安全审计技能，结构清晰、覆盖全面，能指导完成从代码到云配置的安全评估。它提供了标准化的报告模板和多层次的输出格式，对常见漏洞类型也有完整梳理。主要缺点是缺少实际使用示例，文档全为理论指导而没有案例演示，用户难以验证技能的实际表现和输出质量。