control-chrome

Fail

Audited by Socket on Aug 24, 2026

2 alerts found:

MalwareSecurity
MalwareHIGH
agents/openai.yaml

This is a high-risk, capability-expanding agent prompt that explicitly enables unrestricted autonomous Chrome control, suppresses action-time confirmation, instructs inspection of highly sensitive browser secrets (including passwords/session data), and directs bypassing authentication barriers instead of requiring sign-in/verification. It also grants scope to perform potentially privileged operations (deploy and DNS/service changes) via browser automation. Even without explicit exfiltration code in the snippet, the designed behavior strongly aligns with credential/session compromise and unauthorized high-impact actions in a supply-chain/agentic setting.

Confidence: 78%Severity: 98%
SecurityMEDIUM
SKILL.md

SUSPICIOUS. The skill is largely aligned with browser control, but it grants unusually broad authority to use sensitive browser state and to perform consequential actions without per-action confirmation. No clear malicious exfiltration or untrusted installer is present, but the autonomy and credential-adjacent access make it high-risk for misuse.

Confidence: 86%Severity: 78%
Audit Metadata
Analyzed At
Aug 24, 2026, 12:17 AM
Package URL
pkg:socket/skills-sh/samihalawa%2Fopenai-bundled-unrestricted%2Fcontrol-chrome%2F@2f1e8d5cebd286047aecf000fef6b62901eba40ef91cb62e1aa2474830b2f5e8