side-quest

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill requires recording and echoing the user's "verbatim" words into the inbox, receipts, and filenames (and to print confirmations), so any secret the user includes would be reproduced verbatim by the agent—creating an exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.30). The skill captures and forwards user-provided text (triggered by markers like “sq:” / “background this”) into a dispatch brief and writes verbatim words into an inbox, so outsider-authored free text is ingested into the runtime workflow before any “specific item” selection.