The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could serve as a vector for indirect prompt injection attacks.
Ingestion points: The skill uses functions such as read_sbml_model, load_json_model, and load_yaml_model (found in SKILL.md and references/api_quick_reference.md) to ingest user-provided metabolic model files.
Boundary markers: The instructions do not specify any boundary markers or delimiters to isolate content from these files or warn the agent to ignore embedded instructions.
Capability inventory: The skill has access to the Bash tool for package installation, the Write tool for creating CSV, PNG, and XML files, and network access via the cobra library to fetch remote models.
Sanitization: There is no evidence of sanitization or validation performed on the textual content within the model files before they are processed by the agent.
[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands to manage the environment and perform file operations.
Evidence: SKILL.md and references/workflows.md contain uv pip install commands for library installation and instructions for writing analysis results to the filesystem (e.g., to_csv, savefig).
[EXTERNAL_DOWNLOADS]: The skill fetches metabolic models from external scientific repositories.
Evidence: The load_model function in SKILL.md is documented to fetch models from the BiGG and BioModels databases when network access is available. These are well-known and established scientific services.

cobrapy