ascend-migration-analysis

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

• Third-party content exposure detected (high risk: 0.85). Skill runtime ingests outsider-authored free text when it analyzes the target project’s source files (e.g., via rg/cat over requirements*.txt, *.py, and *.cu/*.cuh), which are authored by parties other than the operating user and become readable text in the LLM context through the produced findings/report.