codex-runner
Warn
Audited by Socket on Jul 16, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS: the skill’s main workflow matches its stated purpose, and the underlying Codex CLI is official OpenAI tooling, but the default reliance on a third-party lazycodex/OmO harness plus explicit copying of auth.json tokens into an alternate CODEX_HOME makes the credential and execution footprint broader than a simple bounded worker wrapper. No confirmed exfiltration or malware behavior is present, but the credential forwarding and optional full-access execution raise meaningful security risk.
Confidence: 84%Severity: 61%
Audit Metadata