The Agent Skills Directory

[SAFE]: The skill utilizes well-known, reputable scientific libraries (OpenMM, MDAnalysis, OpenFF, PDBFixer) for its core functionality. All provided code snippets follow standard practices for computational chemistry simulations, including energy minimization and equilibration steps.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external molecular data which represents a standard attack surface for scientific tools.
Ingestion points: Molecular structure files (PDB, SMILES) and simulation trajectories (DCD, XTC) are processed using standard library loaders like PDBFile, mda.Universe, and Molecule.from_smiles in SKILL.md and references/mdanalysis_analysis.md.
Boundary markers: The code relies on the strict parsing logic of the underlying scientific libraries; the agent operating procedure includes instructions to validate results and explicitly confirm parameters with the user.
Capability inventory: Capabilities are limited to simulation execution, data analysis, local file writing for results/logs, and visualization. No arbitrary command execution, dangerous subprocess calls, or unauthorized network access was identified.
Sanitization: The skill includes structural sanitization via PDBFixer to resolve common issues in input PDB files and advises the agent to sanitize molecules before computation.

molecular-dynamics