The Agent Skills Directory

[PROMPT_INJECTION]: The documentation files references/evaluation-frameworks.md and references/system-prompts.md contain examples of prompt injection strings, including instructions to ignore previous rules and 'DAN' jailbreak variants. These are used as test cases for building evaluation frameworks and as examples for implementing prompt defenses, rather than attempts to subvert the agent's actual execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates how to build prompts that ingest untrusted user data (e.g., document summaries, sentiment analysis). While this creates an attack surface for indirect prompt injection, the documentation explicitly provides remediation guidance, including the use of XML tags, instruction hierarchies, and input sandboxing to mitigate these risks.
[EXTERNAL_DOWNLOADS]: The reference guides include examples of installing standard industry libraries such as scikit-learn, tiktoken, anthropic, and openai. These are well-known packages in the prompt engineering and data science ecosystems and do not represent a supply chain risk in this context.
[COMMAND_EXECUTION]: The skill provides examples of GitHub Actions workflows and Python snippets for automated testing. These scripts use standard parameters and environment variables, following secure coding practices for CI/CD integrations.

prompt-engineer