The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes reStructuredText (RST) files from external repositories (e.g., QEMU checkouts), which presents an attack surface where documentation content could be misinterpreted as instructions by the agent.
Ingestion points: Reads .rst files and repository structure information (via find and git) to build or update documentation.
Boundary markers: The instructions do not explicitly provide delimiters or warnings to ignore instructions embedded within the documentation files being processed.
Capability inventory: The skill can perform file system writes (audit logs, scripts), execute shell commands (bash scripts, Ninja, Meson), and install Python packages.
Sanitization: No specific sanitization or escaping mechanisms are described for the content ingested from the external repository.
[DYNAMIC_EXECUTION]: The skill utilizes shell-based heredocs to generate and execute Python code snippets for syntax validation.
Evidence: Both scripts/rst-gates.sh and references/docs-validation.md contain logic that executes Python code via `python3
<<'PY'` to perform docutils parsing and syntax checks.
[EXTERNAL_DOWNLOADS]: The validation workflow instructions suggest downloading and installing common utility packages from the official Python Package Index (PyPI).
Evidence: The references/docs-validation.md file recommends installing docutils, sphinx, and sphinx_rtd_theme using pip to facilitate documentation builds and validation.

qemu-rst-documentation