The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill provides an example fixture for integration testing in SKILL.md that contains a hardcoded email ('admin@example.com') and password ('AdminPass123!'). While clearly intended as placeholders for a local test environment, this pattern can lead to credential exposure if developers copy the code directly into production-facing test suites.\n- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates patterns for reading and processing external data files (e.g., CSV and JSON) within the test environment. If an agent executes these tests on untrusted files, it creates an attack surface for indirect prompt injection.\n
Ingestion points: SKILL.md provides examples of reading files using tmp_path and process_large_file function calls.\n
Boundary markers: No boundary markers or 'ignore' instructions are provided for the data contents in the examples.\n
Capability inventory: The skill leverages httpx for network requests and requires file system access via tmp_path for write/read operations.\n
Sanitization: No data sanitization or validation logic is included in the provided patterns.\n- [COMMAND_EXECUTION]: The skill documents and encourages the execution of various pytest shell commands to run, filter, and monitor tests, which involves executing code within the local environment.

pytest-patterns