reply-writer
Warn
Audited by Snyk on Mar 2, 2026
Risk Level: MEDIUM
Full Analysis
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (high risk: 0.90). The skill explicitly requires ingesting user-generated Reddit content (SKILL.md, Step 1: "Original post — full text of the OP" and "Top existing replies — at least the top 3 replies"), which are public/untrusted third-party inputs the agent must read and use to decide and craft replies, so embedded instructions could influence its behavior.
Audit Metadata