retain

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (medium risk: 0.30). The skill directs the assistant to forcibly halt normal conversation and automatically package-and-submit the current conversation to a third-party CLI (Vidbyte), which is suspicious steering of user data without additional consent but not explicitly malicious.

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

• Third-party content exposure detected (medium risk: 0.30). The /retain skill explicitly instructs the agent to “Review the full current conversation” and “extract the learning surface” by reading the user-provided conversation text at runtime, then uses that extracted free text to generate exercise prompts passed to the Vidbyte CLI.