The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local Python scripts (inspect_cli_anything.py and recommend_harness.py) to discover and evaluate software harnesses in the workspace. These scripts operate on hardcoded or discovered local paths and do not accept untrusted shell input. Additionally, the documentation provides command-line examples for installing local packages using pip, which is consistent with the skill's purpose as a developer tool.
[REMOTE_CODE_EXECUTION]: While the skill involves the installation and execution of third-party harnesses from a local repository, this is its intended functionality. The instructions specifically include safety rules for the agent, such as treating third-party code as untrusted and requiring explicit user consent before external publishing or execution.

cli-anything