The Agent Skills Directory

[SAFE]: The skill is primarily a security utility aimed at providing safe patterns for CLI interactions. Its documentation and examples promote best practices to avoid shell injection attacks.
[COMMAND_EXECUTION]: The skill instructs agents on how to safely assemble and execute commands. The patterns provided, such as using quoted heredocs and avoiding JSON serialization for shell quoting, are industry-standard safety measures.
[DYNAMIC_EXECUTION]: The evaluation harness within the evals/ directory uses Python's subprocess.run and Node.js's vm module to validate code snippets. These operations are restricted to a local, temporary testing environment with limited permissions and no access to external services or credentials.

cli-text