The Agent Skills Directory

[SAFE]: The skill contains no executable code, scripts, or binaries. It is composed entirely of markdown-based instructions for the AI.- [SAFE]: All external URL references are to well-known and trusted sources (Wikipedia), used solely for technical reference on writing patterns.- [SAFE]: The skill's file system access is restricted to reading a local configuration file ('world-code/voice.md') intended for styling instructions, posing no threat of sensitive data exposure.- [PROMPT_INJECTION]: Although the skill processes untrusted user text and reads a local voice file (Ingestion points), it has no associated tools or capabilities such as subprocess execution, network requests, or file-writing (Capability inventory). No explicit delimiters or sanitization are present (Boundary markers/Sanitization), but the lack of executable capabilities renders the risk of indirect prompt injection negligible.

boring-human