security-audit（安全审计）

Security Audit

Standalone workflow for reviewing a web application or API without depending on other skills.

When to Use

Use this skill when:

• auditing a web application or API for security issues
• reviewing authentication, authorization, or session handling
• checking input validation, injection risk, or data exposure
• performing a structured hardening and reporting pass before release

Scope and Safety

• Confirm the target, authorized boundaries, and whether testing is read-only or allows active probing.
• Do not run destructive checks, high-volume fuzzing, or denial-of-service style traffic unless explicitly authorized.
• Record assumptions, environment limits, and access level before starting.
• Prefer reproducible evidence over broad claims.

Audit Workflow

Phase 1: Scope and Reconnaissance

1. Identify the in-scope domains, apps, APIs, jobs, and background workers.
2. Map the attack surface:

- routes and entry points - authentication flows - admin surfaces - file upload or import paths - third-party callbacks and webhooks

1. Note the framework, hosting model, data stores, and external integrations.
2. Build a simple asset inventory before testing.

Phase 2: Baseline Review

1. Check dependency versions, known exposure areas, and obvious misconfigurations.
2. Review environment and deployment assumptions:

- secret handling - logging - debug mode - CORS - cookie flags - security headers

1. Identify areas where automated scanning would be useful, but keep the workflow self-contained: describe the scan you would run and why.

Phase 3: Manual Web Application Testing

Check:

• injection risk in query params, forms, search, filters, uploads, and template rendering
• XSS risk in stored, reflected, and DOM-driven flows
• broken authentication, weak session handling, and insecure password reset flows
• broken access control and IDOR patterns across user, team, and admin scopes
• CSRF exposure on state-changing requests
• path traversal, file handling, and unsafe object access
• sensitive data exposure in UI, logs, client storage, and error messages

For each finding, capture:

• entry point
• required privileges
• reproduction steps
• observed impact
• fix direction

Phase 4: API Security Review

Check:

• endpoint inventory and undocumented routes
• authn/authz coverage on every state-changing endpoint
• tenant isolation and object-level authorization
• rate limiting and abuse resistance
• request validation and schema enforcement
• unsafe defaults in error responses
• webhook signature verification and replay handling

Where useful, describe the exact request variants that should be tested:

• missing auth
• low-privilege auth
• cross-tenant identifiers
• malformed payloads
• boundary values
• repeated requests

Phase 5: Hardening Review

Review:

• password and token lifecycle
• MFA or step-up auth where appropriate
• least-privilege roles and admin separation
• audit logging for privileged actions
• secrets management and rotation
• SSRF, open redirect, and outbound request controls
• file upload validation and storage isolation
• backup, recovery, and incident response readiness

Phase 6: Reporting

Produce a report with:

1. Executive summary
2. Scope and methodology
3. Findings ordered by severity
4. Reproduction notes and evidence
5. Remediation guidance
6. Residual risks and follow-up checks

Security Checklist

Web

• Authentication flows reviewed
• Authorization boundaries reviewed
• Session handling reviewed
• Input validation reviewed
• Injection risk reviewed
• XSS risk reviewed
• CSRF protection reviewed
• Error handling and data exposure reviewed

API

• Endpoint inventory captured
• Auth coverage checked per endpoint
• Object-level authorization checked
• Rate limiting reviewed
• Validation and schema handling reviewed
• Webhook verification reviewed

Operations

• Secrets handling reviewed
• Logging and audit trail reviewed
• Security headers and cookie settings reviewed
• Dependency posture reviewed
• Recovery and incident readiness reviewed

Output Standard

Every finding should include:

• title
• severity
• affected surface
• reproduction steps
• impact
• remediation
• confidence level

Limits

• This skill structures the audit; it does not replace environment-specific testing or specialist review.
• If access, authorization, or rules of engagement are unclear, stop and clarify before continuing.