The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains several examples of prompt injection patterns (e.g., "Ignore previous instructions", "You are now a different AI"). These are correctly categorized and documented as part of an adversarial test suite for evaluating the safety and robustness of an AI pipeline, rather than instructions to be followed by the agent.
[COMMAND_EXECUTION]: Includes examples of SQL commands for maintaining a local results database and shell commands for execution within a CI/CD environment (e.g., npm ci, node scripts/run-evals.js). These are standard development and automation practices.
[EXTERNAL_DOWNLOADS]: References the well-known promptfoo utility and provide a link to its official GitHub repository. This is recommended as a standard tool for systematic LLM testing and quality regression monitoring.

eval-harness