The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill is designed to execute Python scripts on a remote server through the inference.sh platform. It directs the user to install a command-line utility from a public GitHub repository (inference-sh/skills) to facilitate this execution.
[DYNAMIC_EXECUTION]: The primary function of this skill is the dynamic execution of Python code provided as input. While documented as sandboxed, this allows the agent to generate and run logic at runtime, including complex operations like browser automation and video processing.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect injection due to its high-tier capabilities and data ingestion patterns.
Ingestion points: The code field in the input schema within SKILL.md allows for arbitrary instructions to be passed to the executor.
Boundary markers: There are no specific boundary markers or instructions to sanitize the code input against embedded malicious logic.
Capability inventory: The execution environment includes powerful libraries for network access (requests, httpx, aiohttp), browser automation (selenium, playwright), and file manipulation.
Sanitization: The skill relies on the remote platform's sandboxing mechanism for isolation, but does not provide local validation of the code being executed.

python-executor