The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from genomic interval files (BED), sequence files (FASTA), and fragment files (TSV), which presents an attack surface where malicious instructions could be embedded in the processed data.
Ingestion points: Data is ingested via methods like RegionSet.from_bed, RefgetStore.from_fasta, and RegionSet.from_json as documented in SKILL.md and references/python-api.md.
Boundary markers: The skill documentation does not mention the use of delimiters or instructions to the agent to treat the contents of these genomic files as untrusted or to ignore embedded instructions.
Capability inventory: The skill allows the agent to read and write files (e.g., to_bed, to_json, save) and execute various command-line operations for genomic processing.
Sanitization: There is no evidence of content validation or sanitization for instructions embedded within the genomic data formats.
[EXTERNAL_DOWNLOADS]: The skill includes functionality to fetch data from the BEDbase genomic registry.
Evidence: The bbcache module, documented in references/cli.md, uses the gtars bbcache fetch --id <bedbase_id> command to download data from bedbase.org, which is a recognized service for genomic data management.

gtars