The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the use of the belt CLI tool to run Python scripts. These commands are descriptive of the skill's primary functionality for sandboxed execution and do not involve unauthorized privilege escalation or hidden execution.
[EXTERNAL_DOWNLOADS]: The skill references documentation and installation scripts from raw.githubusercontent.com/inference-sh/* and inference.sh, which are the official domains associated with the skill author and service provider. These are considered trusted references for the context of this skill.
[INDIRECT_PROMPT_INJECTION]: As a tool designed to execute arbitrary Python code, the skill naturally has an ingestion surface for code inputs. However, it is explicitly presented as a sandboxed environment for code execution, and the documentation provides clear schemas and examples for intended use.
[SAFE]: No obfuscation, data exfiltration patterns, credential theft, or persistence mechanisms were found. The skill aligns its stated purpose (Python execution) with its documented behavior.

python-executor