The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill generates Python test scripts (tests/smoke/test_NN_*.py) based on project metadata and executes them using pytest. This behavior is the core intended functionality for verifying ML pipeline structural correctness.
[INDIRECT_PROMPT_INJECTION]: The skill reads and processes local files including dataset samples, design journals, and experiment scripts. This creates a vulnerability surface where malicious content in these files could attempt to influence the agent's logic.
Ingestion points: data/, journal/NN_*.md, experiments/NN_*.py, src/<pkg>/data.py.
Boundary markers: None explicitly mentioned in the instructions for isolating external content.
Capability inventory: File system write (tests/smoke/) and command execution (pytest via skore_skills).
Sanitization: No explicit sanitization or validation of the ingested content is described.
[COMMAND_EXECUTION]: Executes local commands through the skore_skills CLI to perform actions like running smoke tests and checking project status. These operations are scoped to the local environment and vendor-specific tooling.

smoke-test-ml-pipeline