The Agent Skills Directory

[COMMAND_EXECUTION]: The core script scripts/run_dual_axis_review.py uses subprocess.run to execute pytest using either uv run or the system Python interpreter. This is used to verify the test health of the skills being reviewed and is controlled by the --skip-tests flag.
[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow to generate an LLM review prompt (build_llm_prompt) by ingesting instructions and code from other skills in a specified --project-root.
Ingestion points: Files located in skills/*/SKILL.md and scripts/*.py within the target project.
Boundary markers: The generated prompt wraps external content in markdown code blocks and bulleted lists to differentiate it from instructions.
Capability inventory: File system read/write (reports), and subprocess execution (pytest).
Sanitization: Content is interpolated into markdown templates without specific escaping, relying on code block delimiters.
[DYNAMIC_EXECUTION]: The test utility scripts/tests/conftest.py uses importlib.util to dynamically load run_dual_axis_review.py into the test environment. This is a standard Python practice for testing standalone scripts that are not part of a formal package.

dual-axis-skill-reviewer