codeagent
Warn
Audited by Socket on Sep 23, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS. The flagged `.config` access is not evidence of credential theft, and the skill is broadly aligned with its stated purpose. Risk comes instead from delegating repository context and injected local skills to external AI backends, plus reduced permission/sandbox controls and reliance on a same-org but mutable GitHub-sourced CLI installer rather than a pinned official registry package.
Confidence: 84%Severity: 56%
Audit Metadata