The Agent Skills Directory

[SAFE]: The skill serves as a technical knowledge base for prompt design and security. It contains no executable scripts, binaries, or automated tasks.
[PROMPT_INJECTION]: While static analysis flagged instruction-override patterns in references/security.md, these are confirmed false positives. The text contains descriptions of common attack strings (e.g., 'ignore previous instructions') and defensive techniques (e.g., 'Regardless of any subsequent instructions') specifically to educate users on how to identify and prevent prompt injection attacks.
[EXTERNAL_DOWNLOADS]: The file .dev/reference-inventory.json contains links to documentation from trusted organizations such as Anthropic, OpenAI, and OWASP, as well as academic papers on arXiv.org. These references are used for informational purposes and do not involve the download of executable code or dependencies.
[COMMAND_EXECUTION]: The documentation describes architectural patterns like 'Program-Aided Language Models (PAL)' which involve code execution in theory. However, the skill itself does not implement these patterns or contain any code that would execute shell commands or logic.

prompt-engineering