The Agent Skills Directory

[SAFE]: The skill performs character normalization (e.g., converting smart quotes to straight quotes) and phrase flagging using standard local tools (bash, grep, sed).
[SAFE]: The scripts/humanize script includes a security-positive feature that automatically removes invisible Unicode characters (such as zero-width spaces and word joiners) which are often used in prompt injection or data exfiltration attacks.
[DYNAMIC_EXECUTION]: The script optionally uses python3 to perform NFKC normalization on input text. The Python code is a static, one-line script dedicated to character encoding and does not process untrusted code or access system resources.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text but lacks the capability to write to the filesystem (outside of stdout) or make network requests, making it a safe pipe for text processing.

humanizer