sales-redreach

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The content reveals explicit support for large-scale automated outreach (unlimited AI replies + a Chrome extension that throttles sending and includes "anti-detection" measures) plus webhook pushes to arbitrary endpoints — features that intentionally enable evasion of platform defenses and mass unsolicited messaging, creating a high risk of abuse.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). Outsider free text can enter the LLM context via the “AI reply suggestions” workflow: Redreach ingests matched Reddit posts/comments (authored by other users) and generates reply drafts using that post text, which is then fed into the agent’s LLM context.