agentmail
Fail
Audited by Socket on Mar 8, 2026
1 alert found:
MalwareMalwareSKILL.md
HIGHMalwareHIGH
SKILL.md
The skill's static findings include a line referencing 'Ignore previous instructions' as part of an example in SKILL.md. This is a potential prompt-injection vector if an agent were to treat the example as an actionable instruction. The presence in documentation does not equate to active execution, but it signals the need for caution in how prompts could be influenced.
Confidence: 85%Severity: 70%
Audit Metadata