voice-extract

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

• Third-party content exposure detected (high risk: 0.90). The skill's required workflow ingests and analyzes user-provided writing samples from the Sample Intake (directory mode copying files from a user-specified folder and interactive paste) and then runs those untrusted/user-generated samples through the SICO extraction (Pass 1/Pass 2) to produce registers that drive later behavior, so public/social-source content (e.g., Reddit/slack examples mentioned) is explicitly consumed and can materially influence agent actions (SKILL.md Step 1 + modules/sample-intake.md and modules/sico-extraction.md).