The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions facilitate the installation of the gget Python package and the download of significant external resources, including AlphaFold2 protein model parameters (~4GB), CELLxGENE single-cell census data, and the Eukaryotic Linear Motif (ELM) database from established scientific repositories.
[COMMAND_EXECUTION]: The skill includes functionality to invoke local bioinformatics binaries such as muscle for multiple sequence alignment and diamond for high-throughput protein alignment through subprocesses.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted genomic sequences and metadata fetched from numerous external databases, creating a potential surface for injection attacks embedded in scientific data. 1. Ingestion points: The skill retrieves data from remote REST and GraphQL APIs (Ensembl, NCBI, UniProt, OpenTargets) and local TSV files (COSMIC) as described in SKILL.md and references/databases_workflows.md. 2. Boundary markers: No explicit instructions were found to delimit or escape sequence data during processing. 3. Capability inventory: The skill possesses capabilities for network operations (fetching data), file system modifications (saving PDB, CSV, and FASTA files), and subprocess execution (alignment tools). 4. Sanitization: No specific sanitization or validation logic for ingested genetic sequences is documented, relying on the default behavior of the gget library.

gget-genomic-databases