The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's primary function is to send user-defined or agent-generated Python code to the vendor's remote environment (inference.sh) for execution. This is the documented purpose of the tool and utilizes the vendor's own infrastructure.
[COMMAND_EXECUTION]: The skill relies on the infsh command-line utility to interact with the vendor's API and execute tasks. The allowed-tools section explicitly permits Bash(infsh *) for this purpose.
[PROMPT_INJECTION]: As a code execution engine, the skill inherently possesses an indirect prompt injection surface. It ingests arbitrary strings into a code execution context.
Ingestion points: The code parameter in the input schema (SKILL.md).
Boundary markers: Not explicitly defined in the prompt template, though the skill description mentions a sandboxed environment.
Capability inventory: Includes full Python execution with network access (via requests/httpx), browser automation (Selenium/Playwright), and file system access within the remote sandbox (writing to outputs/).
Sanitization: The skill relies on the vendor's server-side sandboxing for security rather than input sanitization.

python-executor