geniml
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process genomic interval data and manifest files, creating a potential surface for indirect prompt injection if malicious instructions are embedded in external data inputs.
- Ingestion points: The skill reads local BED files via
scripts/bed_validator.pyand project manifests in CSV/TSV format viascripts/corpus_auditor.py. - Boundary markers: Scripts enforce strict schema validation for delimiters, genomic coordinates, and YAML configuration keys.
- Capability inventory: Included scripts are dependency-free and limited to local file system operations using standard Python libraries; they do not perform network requests or execute subprocesses.
- Sanitization: Robust path validation in
scripts/_common.pyrejects URLs, symlinks, and parent directory traversal to prevent unauthorized file access. - [EXTERNAL_DOWNLOADS]: The skill documentation references external services for package installation and model retrieval.
- Mentions installation of scientific libraries from public package registries.
- Describes the retrieval of model checkpoints and configuration files from the Hugging Face platform.
- References genomic data retrieval from the official BEDbase API.
Audit Metadata