Humanize Chinese

What to consider before installing or running this skill: - Functional fit: The package is local, pure‑Python, and does what it claims: pattern + n‑gram detection and rule/statistics‑guided rewriting. The SKILL.md instructions align with the included scripts. - No hidden credentials or network calls were found in the provided excerpts; the code reads local JSON files (patterns_ngram etc.) and calls local scripts. Still, always review the full scripts (including truncated portions) before running. - Execution risk: The skill needs permission to execute local Python scripts (SKILL.md lists exec). If you allow the agent to invoke skills autonomously, it could run those scripts without further prompts — sandbox or restrict autonomous execution if you don't want that. - Ethical/legal risk: This tool explicitly aims to reduce AIGC detection scores and to '降重' for academic submissions (CNKI/VIP/Wanfang). That is a meaningful misuse/academic‑integrity risk. Even though it's technically coherent, using it to evade institutional checks may violate rules or laws — do not use it to cheat or to conceal authorship. - Practical recommendations: (1) Review the full source locally (especially ngram_model.py and any truncated code) before execution. (2) Run in an isolated environment or container if you are unsure. (3) Use outputs as editing suggestions only — read and verify rewritten text, citations, and logic. (4) Avoid using 'aggressive' modes on sensitive or authoritative documents without manual review. (5) If you want extra assurance, run the scripts on non‑sensitive sample files first to observe behavior. If you want, I can scan the remaining truncated files for network I/O or dangerous system calls and flag any suspicious code paths I find.