The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several command-line tools to perform its primary function, including npx promptfoo for prompt evaluation and pytest for executing DeepEval tests.
[DYNAMIC_EXECUTION]: The skill explicitly runs a local Python script located at evals/deepeval/config.py to facilitate evaluation execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from JSON files located in evals/results/. These files contain the output of AI evaluations, which may include untrusted content from the models being tested.
Ingestion points: Evaluation results in JSON format read from evals/results/ (SKILL.md).
Boundary markers: No specific delimiters or safety instructions are defined for the parsing of these results.
Capability inventory: The skill possesses the ability to execute shell commands (npx, pytest, python) and write files to the local filesystem (evals/results/validation_report.md).
Sanitization: There is no evidence of sanitization or validation of the content within the results JSON files before processing.

evals-validate