The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection, as it processes untrusted input from source files and user descriptions to generate executable code.
Ingestion points: scripts/generate_tests.py reads content from local source files via the --file flag and takes natural language input from the CLI description argument.
Boundary markers: Absent. The untrusted data is interpolated directly into the user_prompt string without delimiters or instructions to the model to ignore embedded commands.
Capability inventory: The skill has the capability to write the resulting LLM output to the local filesystem using Path.write_text in scripts/generate_tests.py.
Sanitization: No sanitization or validation is performed on the source code before it is sent to the LLM, nor on the resulting code before it is written to disk.
[EXTERNAL_DOWNLOADS]: The skill utilizes the official openai Python library and connects to OpenAI's well-known API services to perform its core function. These operations are transparent and align with the skill's documented intent.
[COMMAND_EXECUTION]: The skill provides Python scripts that perform file system operations (reading and writing files). These are restricted to the local workspace and are necessary for generating and managing test files.

testing-framework