skill-reviewer
Warn
Audited by Socket on Jul 12, 2026
1 alert found:
SecuritySecurityevals/fixtures/prompt-injection/SKILL.md
MEDIUMSecurityMEDIUM
evals/fixtures/prompt-injection/SKILL.md
SUSPICIOUS: the stated purpose is calibration, but the skill embeds active prompt-injection instructions plus shell execution and outbound data collection behavior. Even as a test artifact, its actual footprint is disproportionate to the purpose and would expose secrets if followed.
Confidence: 96%Severity: 86%
Audit Metadata