The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze source code within the workspace (such as the provided orders.py module) to generate test suites. This core functionality creates an attack surface for indirect prompt injection where instructions could be embedded in the code being analyzed.
Ingestion points: Source code files and modules provided by the user in the agent's workspace.
Boundary markers: The instructions use standard markdown headers and code blocks but do not specify specialized security delimiters or instructions to neutralize potential commands within processed code.
Capability inventory: The agent is authorized to write files (test_orders.py) and is guided to use standard development tools such as pytest and playwright.
Sanitization: No explicit content sanitization or filtering requirements are defined for the agent when processing workspace files.
[EXTERNAL_DOWNLOADS]: The skill's documentation and evaluation setup reference external development resources, including official Docker images (e.g., postgres:16) for testcontainers and official GitHub Actions (e.g., actions/checkout@v4). These references target established and trusted services within the development ecosystem.
[COMMAND_EXECUTION]: The skill's evaluation harness executes a deterministic grader script (graders/check.py) using Python 3. The script performs static analysis of code structures using the standard library ast module and does not execute the code being graded.
[DYNAMIC_EXECUTION]: The skill package includes a shell script (solutions/reference-test-orders) that generates a reference Python test suite via a heredoc. This is used for internal validation of the grading system and does not involve runtime execution of untrusted remote content.

testing-patterns