bio-causal-genomics-mendelian-randomization

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

• Third-party content exposure detected (medium risk: 0.30). SKILL runtime reads user-provided GWAS summary-statistics files via TwoSampleMR::read_exposure_data/read_outcome_data (e.g., filenames/TSV contents), so outsider-authored free text can flow into the LLM through those datasets.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly calls remotes::install_github(...) for required R packages (e.g., remotes::install_github('MRCIEU/TwoSampleMR')), which fetches and executes remote code at install time and is presented as a required runtime/prerequisite dependency.