meridian-privilege-escalation
Warn
Audited by Socket on Oct 23, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS: the skill is purpose-aligned as a privilege-escalation guide, but its whole function is to help the agent weaken or bypass sandbox and approval boundaries, including switching to less-restricted harnesses. There is no clear exfiltration, credential harvesting, or covert behavior in the text, and broad escalation is framed as requiring user approval. Main risk is operational: if used carelessly, it expands the agent's real-world permissions substantially.
Confidence: 82%Severity: 58%
Audit Metadata