ai-automation-workflows

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md shows runtime workflows that pass user-provided free text (e.g., $INPUT_TEXT from script arguments and $(cat $file) from arbitrary .txt files) into belt app run ... --input, which becomes LLM prompt context; this is outsider-authored content if the operating user supplies/uses files or arguments originating from others.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (medium risk: 0.60). The skill instructs creating persistent cron jobs and writing logs to system paths like /var/log and /output (root-owned locations), which can modify the host state and may require elevated privileges even though it doesn't explicitly ask for sudo.