find-vulnerabilities

SUSPICIOUS: the skill is internally coherent and mostly well-scoped for local code auditing, with official tool references and strong rules against live targets, credential use, and prompt-following from reviewed content. However, it is still a high-risk offensive-security skill for AI agents because it enables vulnerability discovery and exploit-condition testing, so overall risk is elevated despite low evidence of malware or exfiltration.

The code illustrates unsafe deserialization and XML external entity risks in explicitly named vulnerable functions. These functions are dangerous if given untrusted input, but the fragment itself shows no evidence of malicious intent, data exfiltration, or destructive behavior. Prefer fixed_load_session for untrusted serialized data and disable or reject external entities and DTDs for untrusted XML.

This is intentionally vulnerable demonstration code. Its vulnerable mode can disclose the contents of a caller-selected local file through XXE and standard output. The fixed mode blocks the demonstrated DOCTYPE payload. Risk depends on running the vulnerable mode; the fragment does not show network exfiltration or persistence.

The fragment contains an unsafe deserialization vulnerability if vulnerable_load receives untrusted YAML. fixed_load uses the safer API. No evidence of malicious intent or malware is present.

This is a database search helper with SQL injection vulnerabilities when `q` or `sort` is untrusted. Bind `q` as a SQL parameter and restrict `sort` to an allowlist of column names. No evidence of malicious behavior in the fragment.