The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill clones a third-party repository from GitHub (github.com/dauparas/ProteinMPNN.git) and executes a Python script (protein_mpnn_run.py) directly from the cloned source. This behavior constitutes the execution of unverified code from an external source.
[EXTERNAL_DOWNLOADS]: The instructions require downloading external Python packages (torch, numpy) and cloning a remote Git repository to access the model weights and runtime scripts.
[COMMAND_EXECUTION]: The skill uses shell commands, including pip install, git clone, and python, to set up and run the SolubleMPNN tool. These commands are executed in the user's environment to process data and manage dependencies.
[PROMPT_INJECTION]: The skill processes user-supplied protein structure files (.pdb) which serve as complex data inputs for the agent's logic, creating a surface for indirect prompt injection.
Ingestion points: The --pdb_path argument in protein_mpnn_run.py ingests external structure files (e.g., backbone.pdb).
Boundary markers: Absent; there are no instructions provided to the agent to treat the PDB content as untrusted data or to use specific delimiters.
Capability inventory: The skill has the ability to execute shell commands and run downloaded Python scripts.
Sanitization: Absent; the skill relies entirely on the third-party implementation to handle file parsing and does not specify any local sanitization steps.

solublempnn