The Agent Skills Directory

[PROMPT_INJECTION]: Reference files including references/system-prompts.md and references/evaluation-frameworks.md contain strings associated with prompt injection, such as 'Ignore all previous instructions', 'reveal your prompt', and 'You are now DAN'. These markers were evaluated and determined to be part of educational test cases designed to help users build robust evaluation frameworks for testing their own prompts against adversarial inputs. They do not constitute an attempt to hijack the agent's behavior.
[COMMAND_EXECUTION]: Several reference files contain Python code snippets and GitHub Actions workflows. These are provided as static educational examples for tasks like calculating F1 scores, running A/B tests, or automating evaluation pipelines. These snippets are not executed by the skill itself.
[EXTERNAL_DOWNLOADS]: The documentation references standard software development practices, including the use of well-known libraries such as openai, anthropic, and tiktoken. These are standard tools in the domain of prompt engineering and are used here within benign educational contexts.

prompt-engineer