The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external genomic files (BED, GFF, VCF, SAM, etc.) which act as ingestion points for untrusted data into the agent's context.
Ingestion points: Data is read from the filesystem by scripts/audit_intervals.py, scripts/check_contigs.py, scripts/convert_coords.py, and scripts/normalize_variant.py using standard file reading methods.
Boundary markers: The skill does not employ specific delimiters or "ignore instructions" warnings when presenting the output of its genomic processing scripts to the agent.
Capability inventory: The scripts are limited to file reading, arithmetic calculations, and text formatting. They do not utilize high-risk capabilities such as eval(), exec(), or subprocess calls.
Sanitization: The processing scripts enforce strict format validation, such as verifying integer coordinates and ACGTN base sequences, which reduces the surface area for interpreting data as natural language instructions.

genomic-coordinates