po-once-agent-api

Warn

Audited by Socket on Apr 20, 2026

2 alerts found:

AnomalySecurity
AnomalyLOW
scripts/po-once.cjs

No clear evidence of intentional malware (no backdoor, obfuscation, or covert local compromise) is present in this module. The dominant risk is trust boundary weakness: the tool will upload arbitrary user-selected local files to an uploadUrl returned by the remote API and will send the API key to whatever baseUrl is provided by environment/config, with no allowlisting/pinning/validation. Additionally, it stores the API key in plaintext JSON on disk. In a normal deployment with controlled configuration and trusted server responses, this is likely a legitimate CLI client; in hostile or misconfigured environments, it can enable local data exfiltration.

Confidence: 66%Severity: 58%
SecurityMEDIUM
SKILL.md

SUSPICIOUS. The skill's capabilities broadly match its stated social-posting purpose, but trust is weakened because the API key is routed through a non-verifiable Po Once endpoint on generic Convex hosting, and the base URL can be overridden to another host. This is not confirmed malware, but it carries meaningful credential-forwarding and autonomous action risk.

Confidence: 85%Severity: 71%
Audit Metadata
Analyzed At
Apr 20, 2026, 04:56 AM
Package URL
pkg:socket/skills-sh/Stiffjobs%2Fpo-once-agent-skill%2Fpo-once-agent-api%2F@54d61f9e9765f9d8f1c2a0d4798419fdd963538c