The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides examples of prompt injection strings, such as "ignore previous instructions" and "DAN" variants, within sections dedicated to "Injection Test Suite" and "Test Case Format." These are educational resources designed to help developers evaluate and strengthen the security of their own LLM prompts.
[EXTERNAL_DOWNLOADS]: Documentation references standard, reputable software libraries including pydantic, tiktoken, and scikit-learn, as well as official SDKs for major AI platforms. These are listed as recommended tools for building the systems described in the guides.
[COMMAND_EXECUTION]: Example scripts and CI/CD configurations include shell commands for running evaluations and managing environment variables. These are provided as informational templates for the user's external development workflows rather than instructions for the agent to execute.

prompt-engineer