The Agent Skills Directory

[PROMPT_INJECTION]: The files SKILL.md and JAILBREAK_PATTERNS.md contain numerous examples of prompt injection, instruction overrides, and jailbreak payloads (such as DAN, developer mode, and encoding bypasses). While these are explicitly framed as educational reference material and testing guidelines, their presence in the text could cause context confusion or accidental execution by the processing model if strict boundaries are not maintained.
[NO_CODE]: The skill consists entirely of documentation and reference playbooks in markdown format. No executable scripts, binaries, tool configurations, or installation manifests are included, eliminating the risk of direct command execution or malicious supply-chain dependencies.

llm-prompt-injection