The Agent Skills Directory

[SAFE]: No security issues detected. The skill is a standard wrapper for the qwen CLI tool.
[COMMAND_EXECUTION]: The skill executes local commands using a Python wrapper to interface with the qwen CLI. This behavior is the primary intended function of the skill and is explicitly documented for the purpose of executing model prompts.
[INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for untrusted data as it processes user prompts and external files via the --prompt-file argument. This is an inherent property of model runner skills.
Ingestion points: User-provided strings in run_glm.py and file content via the --prompt-file CLI option.
Boundary markers: None explicitly defined in the wrapper script; boundaries are handled by the underlying shared runner implementation.
Capability inventory: Execution of model prompts which may involve tool calls through the qwen CLI.
Sanitization: The skill supports safety-oriented parameters including --safe, --sandbox, and --restrict-tools to constrain the execution environment.

glm-runner