The Agent Skills Directory

[PROMPT_INJECTION]: The documentation includes illustrative examples of prompt injection techniques such as instruction overrides and jailbreaking attempts (e.g., "Ignore all prior instructions", "DAN") used for testing LLM safety boundaries.\n- [PROMPT_INJECTION]: Educational examples demonstrate the use of obfuscation methods including Base64 encoding, zero-width characters, and Unicode homoglyphs to simulate adversarial inputs.\n- [EXTERNAL_DOWNLOADS]: Reference files provide links to established security tools and well-known research organizations (e.g., MITRE, Atomic Red Team, Rapid7, OWASP) for educational and testing purposes.\n- [NO_CODE]: The skill consists entirely of Markdown documentation and contains no executable scripts or automated command patterns.

red-teaming