hermes-client-web-ui

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes explicit plaintext credentials (admin@admin.com / 123456) and examples that embed JWT/Bearer tokens in headers and query strings, which encourages the agent to handle or output secret values verbatim.

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow includes reading outsider-authored free text from uploaded files: the UI/API accepts arbitrary user-provided file content via /api/conversations/:conversationId/upload, stores it, and then passes it into the Hermes chat context (either via --image <absolutePath> for images or by embedding File: <absolutePath>\n\n${message} for other files), so the LLM can ingest attacker-controlled text originating outside the operating user.