The Agent Skills Directory

[Command Execution]: The skill automates development tasks by executing standard tools like uv, ruff, and pytest within the local repository environment. This behavior is expected for its role as a recipe preparation orchestrator.
[Dynamic Code Execution]: During Phase 0 and Phase 7, the skill executes inline Python commands and generated test files. These actions are used to validate repository policies and verify that recipes can be imported correctly without side effects.
[Data Processing (Indirect Prompt Injection)]: The skill ingests content from user-provided recipe directories, which could theoretically contain instructions.
Ingestion points: Files within the specified recipe directory (e.g., manifest.yaml, pyproject.toml, and Python source files).
Boundary markers: The skill relies on structured parsing by specialized scripts rather than explicit prompt delimiters.
Capability inventory: The agent can modify the file system (mkdir, touch), edit code, and execute shell commands (uv, ruff, pytest).
Sanitization: External data is processed by internal scripts (such as extract_env_vars.py and align_pyproject.py) before influencing subsequent pipeline phases.

prepare-python-recipe