The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests sequences and gene lists from external files, which could theoretically contain malicious instructions targeting the agent's logic.
Ingestion points: read_fasta in scripts/batch_sequence_analysis.py and read_gene_list in scripts/enrichment_pipeline.py process user-provided FASTA and CSV/text files.
Boundary markers: The skill does not implement specific boundary markers or instruction-ignoring warnings when processing this data.
Capability inventory: Capabilities include extensive network access to bioinformatics APIs, local file system writes for results, and package installation via gget setup.
Sanitization: The scripts validate biological formats but do not sanitize for potential prompt injection content.
[EXTERNAL_DOWNLOADS]: The skill includes setup commands for downloading large-scale scientific datasets and installing module-specific dependencies.
Evidence: gget setup alphafold downloads approximately 4GB of model parameters, and gget setup cellxgene installs necessary Python packages via pip/uv.
[COMMAND_EXECUTION]: The skill interacts with the local environment to manage dependencies and execute specialized bioinformatics binaries.
Evidence: The gget setup command utilizes package managers (pip or uv) to install software, and modules like muscle and diamond invoke external alignment tools.

gget