research-linkedin

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). This skill is designed to automate scraping of LinkedIn profiles (collecting recent post URLs and creating/updating a signals dataset) and explicitly instructs the user to handle CAPTCHAs/warnings to continue automation, which indicates deliberate platform-evasion and potential data-exfiltration behavior (no obfuscated payloads or hidden execs were found).

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

• Third-party content exposure detected (high risk: 0.85). At runtime, after explicit user confirmation, the skill uses browser automation to navigate to Priority 1 LinkedIn profiles and extract recent post text/URLs from LinkedIn pages (public/outsider-authored free text), which is then ingested into the agent context to create signals.