team-linkedin-profiles

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

• Third-party content exposure detected (high risk: 0.85). The workflow ingests outsider-authored free text/structured content from public web sources at runtime via the Exa “people search” results (and potentially Fiber LinkedIn live-fetch), which are then placed into the agent’s LLM context for filtering/deduplication.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill issues runtime requests to the Gooseworks proxy endpoint (e.g., https://api.gooseworks.ai/v1/proxy/orthogonal/run) and uses the returned brand-dev/Exa/Hunter/Fiber responses to construct context strings and queries that directly control the agent's prompts and are required for the skill to function.