astrobrowse-authenticated-agentic-browser
Warn
Audited by Socket on Oct 2, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS: the core browser-automation capability is coherent with the stated purpose and the documented data flow to official AgentPMT endpoints appears internally consistent, but the skill explicitly tells the agent/user to install additional skills from the broader skills ecosystem. That transitive trust chain raises medium risk, especially because this tool can act inside authenticated accounts and perform real-world actions like posting, updating records, and submitting forms. No direct credential harvesting, hidden execution, or malicious exfiltration is shown in this skill alone.
Confidence: 90%Severity: 56%
Audit Metadata