The Agent Skills Directory

[PROMPT_INJECTION]: Several files, including references/system-prompts.md and references/evaluation-frameworks.md, contain strings matching common prompt injection attacks (e.g., 'Ignore previous instructions', 'reveal your prompt', and 'DAN' jailbreaks). These instances were assessed as benign as they are strictly presented as data for testing frameworks (such as the INJECTION_TESTS Python list) or as examples in JSON test objects to help users evaluate the robustness of their own prompts.
[SAFE]: The skill documentation provides code snippets and templates that use well-known, reputable Python libraries such as scikit-learn, tiktoken, pydantic, and scipy, as well as official SDKs for OpenAI and Anthropic. All usage is consistent with established industry practices for prompt engineering.
[SAFE]: No unauthorized network activity, data exfiltration, or persistence mechanisms were detected. The skill functions as a static knowledge base and template generator for professional LLM development workflows.

prompt-engineer