The Agent Skills Directory

[SAFE]: The skill operates as a local utility for text processing. Analysis of humanize.py and detect.py shows standard file I/O and text manipulation logic using the built-in re and unicodedata modules.
[SAFE]: No network operations, external downloads, or third-party dependencies are detected. All processing is performed locally on user-provided input files using the provided scripts and lexicon.
[SAFE]: The skill includes functionality to detect and strip hidden Unicode characters (tags, zero-width joiners, etc.) that can be used for watermarking or data smuggling. This is a security-positive feature designed to protect the user's privacy and text integrity.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided text files.
Ingestion points: The scripts humanize.py and detect.py read arbitrary text from files passed as command-line arguments (e.g., draft.txt).
Boundary markers: Absent. The input text is treated as raw content for analysis and replacement.
Capability inventory: The scripts perform file reads and writes (via the -o flag in humanize.py), but do not perform network operations or command execution on input data.
Sanitization: Present. humanize.py is designed to sanitize text by removing invisible characters and specific lexical patterns.

li-human