The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is explicitly directed to use docs/pages/quickstart/installation.rst from the repository as the "source of truth" for commands and configuration. This creates an attack surface where a repository could contain malicious instructions in its documentation that the agent would execute during setup.
Ingestion points: docs/pages/quickstart/installation.rst via the Read tool.
Boundary markers: Absent; the skill is instructed to prioritize the document's content over its own built-in instructions.
Capability inventory: Access to Bash (including docker exec, uv sync, and python), and the ability to invoke other skills.
Sanitization: None; the content of the documentation is processed and followed without validation or safety filtering.
[DYNAMIC_EXECUTION]: The skill generates and executes Python code snippets at runtime using the -c flag to verify library imports and environment readiness.
Evidence: Executing .venv/bin/python -c "import isaaclab_arena; print(isaaclab_arena.__file__)" in the host environment and similar commands via docker exec in containers.

setup-arena