verbatim-goal-anchoring
Warn
Audited by Socket on Sep 21, 2026
1 alert found:
SecuritySecuritySKILL.md
MEDIUMSecurityMEDIUM
SKILL.md
SUSPICIOUS. The skill is mostly coherent as a local workflow/documentation aid, with no clear exfiltration or supply-chain abuse in the provided text, but it explicitly installs an anti-stop operating pattern: the agent must always restore a static infinite goal and continue chaining work until repeated empty passes. That autonomy-preserving behavior is disproportionate and makes the skill high-risk even without external network abuse.
Confidence: 86%Severity: 72%
Audit Metadata