phasing-imputation

Pass

Audited by Gen Agent Trust Hub on Jul 5, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill outlines a legitimate scientific workflow for genomic data analysis. No evidence of malicious code, unauthorized network access, or credential harvesting was found.\n- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection through the ingestion of external VCF and metadata files. Evidence chain:\n
  • Ingestion points: VCF genotype data and sample metadata (SKILL.md, references/technical_reference.md).\n
  • Boundary markers: Absent.\n
  • Capability inventory: Use of CLI tools like bcftools and Python libraries like pandas for data processing (SKILL.md).\n
  • Sanitization: Not explicitly defined. Findings in this category are assessed as safe due to the absence of exploitable instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 5, 2026, 03:54 AM
Security Audit — agent-trust-hub — phasing-imputation