ss-learn

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/learning.mjs, the only free-text the agent consumes at runtime is outsider-provided JSON passed via --input /path/to/candidate.json to learning.mjs capture, which is then normalized and privacy-scanned from user-supplied strings without requiring the agent to first select any specific trusted item.