skillopt-sleep
Warn
Audited by Socket on Sep 2, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS. The skill's core behavior matches its stated purpose, and the referenced project/docs are same-org Microsoft resources rather than an obvious third-party lure. However, the scope is materially sensitive: it harvests past agent transcripts, may include secrets, forwards derived content to external model backends/CLIs, and can schedule unattended optimization runs that later modify live CLAUDE.md/SKILL.md. This is proportionate to a self-optimization skill, but it creates meaningful privacy, credential-forwarding, and operational risk. No confirmed malware or covert exfiltration beyond the disclosed provider routing is evident.
Confidence: 88%Severity: 58%
Audit Metadata