pair-agent
Warn
Audited by Socket on May 11, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
SUSPICIOUS. The core pairing behavior is coherent with the stated purpose, and the Bun/ngrok references look consistent with official tooling, but the skill intentionally grants another agent remote browser control, can escalate to cookies/storage/JS access, and includes a much broader gstack preamble than needed for pairing. This looks more like a high-risk remote access skill than malware, with the main concern being powerful browser delegation and external tunnel exposure rather than hidden credential theft.
Confidence: 84%Severity: 78%
Audit Metadata