linkedin-browser-providers
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides a framework for managing browser provider sessions without implementing dangerous code or commands. It includes specific security constraints that prevent the agent from requesting, printing, or storing sensitive provider keys or CDP URLs. The use of well-known browser services (Browserbase and Anchor) is handled via environment variables and sanitized identifiers, following standard security practices.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves operating browser sessions on LinkedIn, which exposes the agent to untrusted external data that could contain malicious instructions.
- Ingestion points: Live browser sessions on LinkedIn profiles, messages, or feeds (SKILL.md).
- Boundary markers: The skill instructions explicitly demand that the agent "report only sanitized identifiers" and never print CDP URLs (agents/openai.yaml, SKILL.md).
- Capability inventory: The skill utilizes
doctorandbrowser_sessiontools to manage browser lifecycle and task execution (SKILL.md). - Sanitization: Instructions are provided to ensure that only processed, sanitized identifiers are returned to the user, reducing the risk of data leakage or injection-prompted exfiltration.
Audit Metadata