The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to search for and download force field parameters and interatomic potential files from external literature and databases. This creates a surface where malicious content on a website could influence the agent's behavior.
Ingestion points: The skill uses WebSearch and WebFetch to retrieve parameters and supplementary materials from literature and the NIST Interatomic Potentials Repository.
Boundary markers: There are no specific instructions or markers to help the agent distinguish between simulation parameters and potential malicious instructions embedded in fetched text.
Capability inventory: The skill uses Bash to execute the simulation binary ($LMP) and Write/Edit to generate the input scripts that control the binary's execution.
Sanitization: No sanitization or verification mechanism is described for the data retrieved via the web tools before it is used to generate input files.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute the LAMMPS binary ($LMP). While this is the intended purpose of the skill, the command line is constructed dynamically using a variable that defaults to lmp but can be overridden by the environment.
[EXTERNAL_DOWNLOADS]: The skill fetches interatomic potential files (such as .eam.alloy or .eam.fs) from the NIST Interatomic Potentials Repository. This repository is a well-known scientific resource, and these downloads are documented as standard operational steps for molecular dynamics simulations.

lammps-simulation