zenbin

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt instructs agents to place secret values (API tokens, passwords, URL tokens) directly into request bodies/headers (e.g., auth.credentials, password fields, example snippets), which requires the LLM to handle and emit secrets verbatim.

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

• Third-party content exposure detected (high risk: 0.90). The skill is a public pastebin (ZenBin) that hosts arbitrary user-provided HTML pages accessible at https://zenbin.onrender.com/p/{id} and https://zenbin.onrender.com/p/{id}/raw (and supports external resources and a proxy), so the agent can fetch and interpret untrusted, user-generated web content that could carry indirect prompt-injection payloads.