The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection by design, as it ingests and processes untrusted local data to perform its primary function.
Ingestion points: The skill explicitly reads and analyzes the content of pyproject.toml, pytest.ini, setup.cfg, and all Python files within the tests/ directory (identified in SKILL.md and troubleshooting.md).
Boundary markers: There are no implementation details showing the use of delimiters or instructions to the model to ignore embedded commands within the analyzed files.
Capability inventory: The agent has the capability to execute shell commands (e.g., pip, pytest) and perform file system operations.
Sanitization: No sanitization or validation of the ingested file content is performed before it is used to influence the agent's output and recommendations.
[EXTERNAL_DOWNLOADS]: The skill describes logic to fetch documentation and version information from external sources.
Evidence: Files SKILL.md and async-testing-guide.md contain logic for fetch_latest_docs and get_library_docs, targeting well-known technology projects such as pytest, fastapi (tiangolo/fastapi), and cpython (python/cpython).
Context: These references target well-known services and official repositories for the purpose of providing up-to-date testing guidance.

python-expert-tester