The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for the agent to load and process genome-scale metabolic models from various external file formats.
Ingestion points: The skill utilizes cobra.io functions (read_sbml_model, load_json_model, load_yaml_model, load_matlab_model) to ingest data from external files in SKILL.md and references/api_quick_reference.md.
Boundary markers: The instructions lack explicit boundary markers or directives to the agent to treat content within model files (such as reaction names, metabolite metadata, or gene identifiers) as untrusted data.
Capability inventory: The skill documents capabilities for writing files to the local system (write_sbml_model, to_csv, savefig), performing parallel computations via the processes parameter in functions like double_gene_deletion and sample, and visualizing data using matplotlib and seaborn.
Sanitization: There is no evidence of sanitization or validation of the input model structure or identifiers before processing, which could allow a maliciously crafted model file to influence the agent's reasoning or analysis through embedded instructions.

cobrapy