The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's primary purpose is the execution of arbitrary Python code. It provides an interface to send code strings to a remote backend (infsh/python-executor) via the belt CLI.
[EXTERNAL_DOWNLOADS]: The skill directs users to an external installation script for its required CLI tool located at https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md. This is a remote resource from a source not explicitly defined as a trusted provider.
[COMMAND_EXECUTION]: The skill utilizes the belt CLI tool to perform network-based application runs, which involves executing shell-level commands to interact with the remote executor.
[INDIRECT_PROMPT_INJECTION]: The skill creates a high-capability execution surface. It facilitates web scraping and API interactions (via libraries like requests, selenium, and beautifulsoup4) and then provides a tool to execute Python code. There is a risk that data ingested from external websites could influence the generation of Python code, leading to unintended code execution if the agent does not strictly validate inputs.
Ingestion points: External data enters the context through web scraping examples and API calls defined in the library list.
Boundary markers: None; the skill provides no delimiters or instructions to ignore embedded code found in scraped content.
Capability inventory: Full Python execution, network access (requests), file system writes (outputs/ directory), and browser automation (playwright).
Sanitization: None; the skill passes raw strings to the executor.
[DYNAMIC_EXECUTION]: The tool is designed to dynamically assemble and execute Python code at runtime. It explicitly supports a wide range of processing tasks by executing code passed through the code parameter in the input schema.

python-executor