The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill generates templates that ingest external data via the datasets library without providing boundary markers or sanitization logic to separate instructions from data.
Ingestion points: env.py (via datasets.Dataset.from_list ingestion into the vf.ToolEnv dataset).
Boundary markers: Absent in the generated create_verifiers_env templates; there are no instructions to the agent to wrap external data in delimiters.
Capability inventory: The generated environment supports stateful sandboxes (e.g., E2B), file system interaction via toolkit methods, and LLM tool calling capabilities.
Sanitization: No sanitization, escaping, or filtering is prescribed for the input dataset content before it is processed by the agent in the environment rollout.
[EXTERNAL_DOWNLOADS]: The skill configuration and generated pyproject.toml include dependencies on external libraries and infrastructure.
Packages: Fetches verifiers, e2b-desktop, openai, and datasets from official registries.
Context: These references target well-known technology providers and official package repositories for the intended development purpose.

generate-verifiers-env