skill-creator-advance

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text is ingested into the LLM context via the runtime eval loop: scripts/run_eval.py writes the user-provided eval_set[].query into the claude -p prompt (and run_loop.py passes those queries through), so any outsider-authored text included in those queries becomes LLM-readable context.