The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions require cloning a repository from an external, third-party GitHub account (https://github.com/hgbrian/biomodals).
[REMOTE_CODE_EXECUTION]: The skill executes a Python script (modal_protenix.py) sourced from the external repository. The execution is handled via the uv tool and the modal cloud platform, which involves running code in a remote environment.
[COMMAND_EXECUTION]: The skill provides shell commands for setting up the environment, creating local input files (e.g., printf ... > target.faa), and launching the external execution process.
[PROMPT_INJECTION]: The skill ingests untrusted data from external biological sequence files, which presents a surface for indirect prompt injection.
Ingestion points: Files provided via the --input-faa or --input-json parameters.
Boundary markers: No delimiters or instructions are provided to the agent to disregard potential instructions embedded within the sequence data.
Capability inventory: The skill executes shell commands and utilizes remote cloud execution capabilities (modal run), which could be abused if the processed data contains malicious directives.
Sanitization: There are no explicit validation or sanitization steps documented for the input data.

protenix