unknown-vector-analysis

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). This skill explicitly instructs the agent to read environment variables and files like /proc//environ and to save/output "evidence" and summaries without guidance to redact sensitive values, so the LLM would likely be required to handle and may output secret values verbatim.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The skill’s workflow includes “WebFetch to retrieve PoC source code from GitHub/exploit-db/security advisories” and then “Inline source code in heredocs is fine,” meaning outsider-authored free text (public PoC/advisory content) can be ingested into the agent’s LLM context via fetched web content at runtime.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs using WebFetch (e.g., WebFetch("")) to retrieve PoC source code from remote sites such as GitHub/exploit-db during runtime and then save/adapt and potentially execute that code, so remote URLs fetched at runtime can directly provide executable instructions.