superhuman
Warn
Audited by Socket on Oct 18, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
SUSPICIOUS: the skill’s email capabilities fit its purpose, but install/execution trust and data handling do not. It depends on a non-official third-party CLI tied to a personal repo, instructs enabling remote debugging on Superhuman, extracts OAuth tokens from the app, caches refresh tokens locally, and then performs real-world email actions on the user’s behalf. Data flows mostly go to expected provider APIs, but the credential extraction and forwarding through third-party code make this a high security concern rather than clearly malicious.
Confidence: 89%Severity: 76%
Audit Metadata