The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides tools for processing external and untrusted data related to molecular and protein modeling, which could potentially contain malicious instructions.
Ingestion points: The skill ingests SMILES strings and protein structures (PDB files) via methods like data.Molecule.from_smiles() and data.Protein.from_pdb() described in references/core_concepts.md and references/protein_modeling.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard instructions that might be embedded within the processed data.
Capability inventory: The skill supports package installations (pip install), filesystem writes (Protein.to_pdb), and network access for downloading datasets (e.g., datasets.BBBP).
Sanitization: No evidence of input validation, sanitization, or safety filtering of the biological/chemical data inputs was found.
[DYNAMIC_EXECUTION]: The skill package contains an executable Python script (translate_md.py) that performs runtime file modifications.
Evidence: The script iterates through a hardcoded list of markdown files, identifies Chinese characters using regex, and overwrites the files with translated content using open(file, 'w').
Evidence: The script includes hardcoded absolute Windows paths (e.g., C:\Users\29707\Desktop\skil11\torchdrug\SKILL.md) that reveal the developer's local username and file hierarchy.

torchdrug-english