DAST Scanner Skill

Perform Dynamic Application Security Testing against a running web application. This skill is language-agnostic -- it tests HTTP endpoints regardless of backend technology (Node.js, Python, Java, PHP, Go, Ruby, .NET, etc.).

Workflow

Step 1: Endpoint Discovery

Identify all available endpoints through multiple strategies:

1. Source code analysis: Read route definition files to extract endpoints.
   • Look for framework-specific routing patterns (Express app.get(), Flask @app.route(), Spring @RequestMapping, Laravel Route::get(), Gin r.GET(), Rails resources, ASP.NET [HttpGet], etc.)
   • Search for files: routes/*.js, urls.py, *Controller.*, router.*, api.*, openapi.*, swagger.*
2. OpenAPI/Swagger specs: Use the Bash tool to fetch /swagger.json, /openapi.json, /api-docs, /swagger-ui via curl if the server is running.
3. Index endpoint probing: GET / and parse response for links and API references.
4. Common path fuzzing: Try well-known paths (/api, /admin, /login, /health, /debug, /status, /graphql, /metrics, /env, /.env, /config, /robots.txt, /sitemap.xml).
5. HTTP method fuzzing: For each discovered endpoint, test OPTIONS to enumerate allowed methods, then try GET, POST, PUT, PATCH, DELETE, HEAD.

Compile a complete endpoint map with: path, supported methods, expected parameters (query, body, header), and authentication requirements.

Step 2: Payload Selection

Based on endpoint characteristics, use the Read tool to load the appropriate testing references:

Step 3: Attack Execution

For each endpoint, use the Bash tool to send crafted payloads via curl. Follow these principles:

• Multiple injection points: Test query parameters, request body (JSON and form-encoded), HTTP headers (Host, Referer, User-Agent, X-Forwarded-For), and cookies.
• Timing-based detection: For blind vulnerabilities, use time-delay payloads and measure response time differences (e.g., time curl ... or compare against baseline).
• Authentication states: Test each endpoint both unauthenticated and with valid credentials (if available). Test with other users' credentials for access control issues.
• Encoding variations: Try raw payloads, URL-encoded, double-encoded, and Unicode variations.
• Safe testing: Prefer read-only probes. Avoid destructive operations (DELETE on real data, DROP TABLE). Use conditional payloads that detect without destroying (e.g., SELECT not DELETE, id not rm).

Use the Bash tool to execute curl with these standard flags:

# Standard request with full response details
curl -s -w "\nHTTP_CODE:%{http_code}\nTIME_TOTAL:%{time_total}\n" -D - "http://TARGET/endpoint"

# POST with JSON body
curl -s -X POST -H "Content-Type: application/json" -d '{"key":"PAYLOAD"}' "http://TARGET/endpoint"

# With authentication
curl -s -H "Authorization: Bearer TOKEN" "http://TARGET/endpoint"

# Timing measurement for blind testing
curl -s -o /dev/null -w "%{time_total}" "http://TARGET/endpoint?param=PAYLOAD"

Step 4: Response Analysis

Analyze HTTP responses for vulnerability indicators:

• Error messages revealing internals: Stack traces, SQL error messages, file paths, framework versions, debug output
• Unexpected data in responses: Database records from other users, internal IPs, configuration values
• Status code anomalies: 200 where 401/403 expected (broken access control), 500 with detailed errors (information disclosure)
• Timing differences: Significantly slower responses may indicate blind SQL injection or SSRF connecting to internal services
• Payload reflection: Input reflected in response without sanitization (XSS indicators, though DAST focuses on server-side)
• Missing security headers: Absence of CSP, HSTS, X-Frame-Options, etc.
• Data leakage patterns: Sensitive data (passwords, tokens, PII) in responses that should not contain them

Step 5: Report Generation

Produce a structured report. For each finding include:

1. Severity: Critical / High / Medium / Low / Info
2. Title: Brief description of the vulnerability
3. OWASP Category: Map to OWASP Top 10 2021 (A01-A10)
4. Endpoint: Exact URL, HTTP method
5. Payload: The exact request that triggered the vulnerability
6. Evidence: The response (or relevant portion) proving the vulnerability exists
7. Impact: What an attacker could achieve by exploiting this
8. Remediation: Specific, actionable fix recommendations appropriate to the detected backend technology

Format the report as a markdown table summary followed by detailed findings sections.

Severity Classification

Key Principles

• Evidence-driven: Every finding MUST include the exact curl command used and the relevant response excerpt proving the vulnerability. No theoretical or assumed vulnerabilities.
• False positive awareness: Distinguish between actual vulnerabilities and expected error responses. A 400 Bad Request with a generic error message is NOT a vulnerability. Verify findings with multiple payloads when uncertain.
• Safe testing: Never execute destructive operations against production data. Prefer SELECT over DELETE, id over rm -rf, sleep over shutdown. Always consider the impact of test payloads.
• Authentication-aware: Test both unauthenticated and authenticated access patterns. When credentials are available, test horizontal and vertical privilege boundaries.
• Language-agnostic: All testing is performed at the HTTP level. Payloads and detection patterns cover multiple backend technologies. Do not assume any specific server-side language.