The Agent Skills Directory

[PROMPT_INJECTION]: The reference files contain educational examples of prompt injection attempts, such as "Ignore previous instructions" and "revealing system prompts". These are explicitly documented as adversarial test cases for evaluating prompt robustness and as examples of what to mitigate when building system guardrails.
[EXTERNAL_DOWNLOADS]: The skill's reference guides include Python code snippets that mention well-known libraries such as scikit-learn, evaluate, tiktoken, anthropic, openai, and pydantic. These are standard tools in the LLM ecosystem and are provided for user implementation rather than automated execution by the agent.

prompt-engineer