The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes metabolic models loaded from external files, which serves as a potential surface for indirect prompt injection if the model data contains malicious instructions.
Ingestion points: Functions such as read_sbml_model, load_json_model, and load_yaml_model in SKILL.md are used to bring external file content into the agent's context.
Boundary markers: No delimiters or specific instructions are defined to separate the model data from the agent's operational instructions.
Capability inventory: The skill possesses capabilities to perform complex simulations and write output files to the system via the COBRApy library.
Sanitization: There is no evidence of validation or sanitization of the loaded model contents before processing.
[SAFE]: The skill provides links to its official documentation on ReadTheDocs and to the author's research platform at k-dense.ai, both of which are treated as legitimate and neutral references.

cobrapy