browser-agent

[Skill Scanner] Installation of third-party script detected This skill documentation describes powerful and legitimate browser automation capabilities consistent with its stated purpose. There is a moderate security risk primarily from operational misuse or misconfiguration: handling of user credentials, potential forwarding of scraped-sensitive data to third-party LLMs, and explicit emphasis on bypassing anti-bot protections. There is no evidence in the provided text of obfuscation or embedded malware; however, because browser automation can perform account actions and exfiltrate data, reviewers should inspect the actual implementation and actionbook recipes for: where credentials are accepted/stored, whether data is sent to external endpoints (LLMs or other domains) without redaction, and whether recipes include third-party intermediaries. Recommended precautions: treat credentials carefully (use ephemeral/test accounts), audit recipe files for external network calls, and configure LLM integrations to avoid sending secrets or PII. LLM verification: The provided SKILL.md (documentation) does not contain direct evidence of malicious code or intentional exfiltration, but it reveals multiple supply-chain and operational risks: unpinned installation instructions (pip install browser-use), absence of secret-handling guidance, and distribution of precomputed automation recipes without demonstrated vetting. Before trusting runtime behavior, review implementation files for network endpoints, telemetry, recipe contents, and any dynamic code executio