sales-sybill

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The prompt embeds bearer-token examples (e.g., "Authorization: Bearer sk_live_...") and instructs verification using a "dashboard signing secret" for webhooks, which encourages the LLM to produce or handle verbatim secrets in headers/commands or code samples.

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

• Third-party content exposure detected (high risk: 0.78). The required runtime workflow for this skill includes Pro+ webhook automations (meeting.new_recording.v1) where Sybill sends payloads containing full transcript and summaries into the agent’s context; those transcripts are outsider-authored free text from meeting participants (not chosen by the operating user), creating an indirect prompt-injection path via webhook-ingested transcript text.