| name | security-audit |
| description | Defensively scan a codebase (or a specific path) for OWASP-style vulnerabilities — injection, path traversal, unsafe reflection/deserialization, SSRF, hardcoded secrets, and XXE — and report each confirmed finding with its file:line location, severity, and concrete remediation. Use when asked to do a security audit, review code for vulnerabilities, find injection/SSRF/secret-exposure risks, or harden a project before release. |
| argument-hint | ["path to scan","optional"] |
| allowed-tools | Read, Glob, Grep, Bash |
Security Audit
Try it: /security-audit:security-audit src/main/java — or say "do a security audit of the auth module". The path argument is optional.
A defensive security review. The goal is to find and report vulnerabilities so they can be fixed — never to exploit them. Produce confirmed findings with file:line, severity, and remediation. Do not generate exploit payloads, weaponized proof-of-concept code, or instructions for attacking a live system.
This skill is language-agnostic. Patterns below use Java / Spring as worked examples; adapt the regexes to the languages actually present in the repo (check file extensions, build files, and framework imports first).
Scope
- If
$ARGUMENTS is provided, limit the scan to that file or directory.
- Otherwise scan all source code in the repository.
- Exclude generated output, vendored dependencies, and build artifacts (e.g.
target/, build/, dist/, node_modules/, vendor/, .git/).
- For secret-style checks, also distinguish production code from test fixtures and report them separately rather than ignoring tests entirely.
Method
For each category: grep for the candidate patterns, then read each hit in context before reporting. A match is only a finding if untrusted input can actually reach the sink without adequate validation, encoding, or sandboxing. Trace data flow far enough to be confident. Report confirmed issues, not theoretical ones.
Checks to Perform
1. Injection (SQL / OS command / template / LDAP)
String-concatenated queries or commands built from user input.
- SQL:
Statement, createQuery(, executeQuery( / executeUpdate( with + string building; raw queries interpolating request data.
- OS command:
Runtime.getRuntime, ProcessBuilder, \.exec\(, plus shell helpers in other languages (os.system, subprocess, child_process, eval, backticks).
- Generic regex:
(createQuery|executeQuery|exec|system|popen)\s*\(.*\+
Remediation: parameterized queries / prepared statements, argument arrays (no shell), allow-lists. Flag any string concatenation of untrusted input into an interpreter.
2. Path Traversal
File or resource paths built from untrusted input.
- Grep:
new File\(.*\b(get(Name|Parameter|Header)|input|request|param), Path\.of\(.*input, Paths\.get\(.*input, \.\./, FileInputStream\(.*request
Verify each hit canonicalizes the resolved path and confirms it stays within an intended base directory (e.g. getCanonicalPath().startsWith(base)), or rejects .. / absolute paths.
Remediation: canonicalize and bound-check against an allowed root; reject path separators in user-supplied filenames.
3. Unsafe Reflection / Deserialization
Dynamic type loading or object deserialization driven by input.
- Reflection:
Class\.forName\(.*input, getMethod\(.*input, \.newInstance\(, setAccessible\(true\)
- Deserialization:
ObjectInputStream, readObject\(, new Yaml\(\), Yaml\.load\( (unsafe vs loadAs/SafeConstructor), readValue\( on polymorphic types with default typing enabled, XMLDecoder, pickle.loads, yaml.load (Python).
Remediation: avoid native deserialization of untrusted data; use type allow-lists, safe parsers (SnakeYAML SafeConstructor, Jackson without enableDefaultTyping), or schema-validated JSON instead. Never reflect into a class name supplied by a caller.
4. Server-Side Request Forgery (SSRF)
Outbound requests whose URL/host comes from input.
- Grep:
(HttpClient|RestTemplate|WebClient|URL\(|openConnection|requests\.get|fetch\(|HttpURLConnection).*\b(input|param|url|request|host)\b
Verify the target URL is validated against an allow-list of hosts/schemes and that redirects and internal/metadata addresses (169.254.169.254, localhost, RFC1918 ranges) are blocked.
Remediation: allow-list destinations, resolve-and-validate the IP, disable automatic redirects, block link-local/private ranges.
5. Hardcoded Secrets
Credentials, tokens, or keys embedded in source or config.
- Grep (case-insensitive):
(password|passwd|secret|token|api[_-]?key|access[_-]?key|private[_-]?key|client[_-]?secret)\s*[:=]
- Also:
-----BEGIN .*PRIVATE KEY-----, AWS-style AKIA[0-9A-Z]{16}, long base64/hex literals assigned to credential-named variables.
Distinguish real values from placeholders (changeme, ${ENV_VAR}, ***) and from test fixtures. Report production secrets as higher severity than test ones.
Remediation: move secrets to environment variables or a secrets manager; rotate any exposed credential; add the pattern to secret-scanning / pre-commit hooks.
6. XML External Entity (XXE)
XML parsing without external-entity processing disabled.
- Grep:
DocumentBuilderFactory, SAXParserFactory, XMLReader, TransformerFactory, XMLInputFactory, SAXBuilder, Unmarshaller
For each, confirm secure configuration: setFeature("http://apache.org/xml/features/disallow-doctype-decl", true), external general/parameter entities disabled, XMLConstants.FEATURE_SECURE_PROCESSING enabled, or IS_SUPPORTING_EXTERNAL_ENTITIES=false.
Remediation: disable DOCTYPE declarations and external entities on every XML parser factory.
Additional: TLS / Certificate Validation Bypass
- Grep:
TrustManager, checkServerTrusted, HostnameVerifier, setHostnameVerifier, verify\(.*return true, InsecureTrustManager, --ignore.*ssl, verify=False
Flag any global disabling of certificate or hostname validation. If a "skip TLS verification" option exists, verify it is opt-in, scoped to a single client, and never affects the global SSL context.
Report Format
Output a single table, ordered by severity (CRITICAL → HIGH → MEDIUM → LOW):
| Severity | Location | Category | Issue | Remediation |
|---|
| CRITICAL | path/to/File.java:42 | Injection | Concrete description of the data flow | Specific fix |
After the table, add a one-line summary (counts by severity).
Report only confirmed findings with a real, traceable data flow — not theoretical risks. If nothing is found, state: "No vulnerabilities found."
