follow-up-after-call

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md requires ingesting “grounded call evidence” from meeting_notes, document_store, or external_messaging connectors, which at runtime can include outsider-authored free text (e.g., customer/partner messages or transcripts) that the agent then uses to draft the follow-up.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill's mandatory pre-answer gate instructs the runtime to load and run the sales:user-context preflight script from ../user-context/SKILL.md, meaning external content at ../user-context/SKILL.md is fetched and its preflight output directly controls agent guidance.