The Agent Skills Directory

[SAFE]: The skill provides legitimate instructions and documentation for using the cuopt_cli utility. The provided examples and sample MPS (Mathematical Programming System) files are consistent with the skill's stated purpose of solving Linear Programming (LP), Mixed-Integer Linear Programming (MILP), and Quadratic Programming (QP) problems.
[COMMAND_EXECUTION]: The skill documents the use of the cuopt_cli command. This execution is scoped to the primary functionality of the skill (numerical optimization) and does not involve unauthorized or dangerous shell operations. Usage of common flags like --time-limit and --mip-relative-tolerance is appropriate for the domain.
[EXTERNAL_DOWNLOADS]: The skill references the cuopt Python package and provides installation instructions using standard package managers (pip, conda). As the package is a legitimate resource from a well-known vendor, this is considered a safe and standard development practice.
[PROMPT_INJECTION]: The skill was evaluated for indirect prompt injection risks since it instructs the agent to analyze the content of external MPS files.
Ingestion points: The agent reads user-provided or asset-resident .mps files to confirm the problem formulation.
Boundary markers: None explicitly defined in the instructions.
Capability inventory: The agent is authorized to execute the cuopt_cli command with parameters derived from the file analysis.
Sanitization: No specific sanitization instructions are present, but the target application (cuopt_cli) is a structured numerical solver, and no patterns suggesting the circumvention of agent safety guardrails were found. The risk is assessed as safe.

cuopt-numerical-optimization-api-cli