The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to configure a pyproject.toml file that includes dependencies on well-known packages such as verifiers, openai, datasets, and e2b-desktop from official package registries.
[COMMAND_EXECUTION]: Includes validation gates that suggest executing shell commands via uv run to test the environment's initialization and rollout logic locally.
[SAFE]: Uses the Python inspect module to dynamically introspect function signatures and docstrings to generate tool schemas, which is a standard and secure approach for implementing tool-calling capabilities in AI agents.
[SAFE]: The skill creates environments that process external data from datasets and execution trajectories. It includes best-practice advice for managing this data, such as truncating large output strings to prevent context overflow and using lazy initialization for sandbox environments.

generate-verifiers-env