bio-atac-seq-enhancer-gene-linking
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements enhancer-gene linking workflows using ABC and ENCODE-rE2G, which are standard methodologies in the genomics community. The scripts provided (e.g., run_abc.sh) utilize common bioinformatics tools like samtools, bedtools, and deepTools for data transformation and analysis.
- [SAFE]: External resources and software are downloaded from official and well-known repositories, including the Broad Institute and Engreitz Lab on GitHub. These represent the primary maintainers of the referenced algorithms.
- [SAFE]: Command execution is constrained to processing scientific data files (BAM, BED, bigWig, Hi-C matrices). There are no attempts to access sensitive system files, establish unauthorized network connections, or implement persistence mechanisms.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external genomic data which serves as a potential ingestion point. However, the risk is negligible as the data formats (binary BAMs, structured TSVs) do not facilitate natural language instruction injection. The skill relies on specialized parsing tools that strictly validate these file schemas.
- Ingestion points: ATAC-seq BAMs, H3K27ac BAMs, and Hi-C contact matrices in SKILL.md and run_abc.sh.
- Boundary markers: Not explicitly defined in the prompt instructions, but enforced by the underlying bioinformatics tools' file format requirements.
- Capability inventory: Execution of local python scripts, shell commands (samtools, bedtools), and file system writes for results.
- Sanitization: Inherited from the input validation of the specialized CLI tools used in the pipeline.
Audit Metadata