agency-request-intake-router

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

• Third-party content exposure detected (high risk: 0.70). The required workflow ingests the user-provided client/internal request text (e.g., email/Slack/PDF content) directly for classification, but that text is authored by the operating user’s client/internal parties; since the skill does not fetch public web content or process outsider-authored documents beyond the provided request, the only free text entering the LLM context is the request itself, which is outsider-authored in typical use (client messages/briefs).

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)