The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill implements strict safety protocols that explicitly prohibit the use of real secrets, live credentials, or customer data. It mandates the use of synthetic data and local-isolated dependencies by default to prevent accidental exposure.
[COMMAND_EXECUTION]: The skill is designed to run project-native test runners such as pytest and unittest. This execution is the intended primary purpose of the skill for verifying code behavior and contract invariants.
[INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risk of processing untrusted data from repository files and test outputs. It incorporates a robust evidence chain for safety:
Ingestion points: Reads repository instructions, existing tests, and generated outputs (SKILL.md, Workflow Step 2).
Boundary markers: Explicitly instructs the agent to treat all repository content and generated output as untrusted data.
Capability inventory: Uses subprocess calls to execute standard Python test runners.
Sanitization: Mandates redaction of secrets, use of project-relative paths, and synthetic data defaults to mitigate potential injection effects.
[SAFE]: The included utility script scripts/plan_case_matrix.py is a specialized tool for calculating test matrices. It uses only Python standard library modules (e.g., argparse, json, random) and lacks capabilities for network access, dynamic code execution, or sensitive file system operations.

python-parameterized-testing