Pro Zh Summary

专业级中文长文本摘要工具。当用户要求“总结”、“摘要”、“提炼要点”、“缩写”或处理超长中文文本时，必须调用此工具进行自动分段与压缩。

This skill appears coherent with its description, but check the following before installing: (1) It requires heavy ML dependencies (transformers, torch, sentencepiece) plus fastapi/uvicorn/requests — the SKILL.md package list omits some of these; install requirements.txt in an isolated environment. (2) On first run server.py will download the model "heack/HeackMT5-ZhSum100k" from the Hugging Face hub and cache it to disk — verify you trust that model repo and expect the network and storage usage. (3) The skill launches a local background process (127.0.0.1:28199) to serve summaries; make sure this behavior is acceptable and run it in a disposable/isolated environment if you have privacy or resource concerns. (4) If you need guarantees about the model's provenance or license, inspect the model repository before use. Otherwise the bundle is internally consistent and matches its stated purpose.

Type: OpenClaw Skill Name: pro-zh-summary Version: 1.1.0 The skill is a legitimate Chinese text summarization tool using a local MT5 model. It employs a client-server architecture where `main.py` acts as a CLI and `server.py` hosts a FastAPI-based inference engine on a local port (28199) to keep the model in memory for performance. While it uses `subprocess.Popen` to silently launch the backend server and allows reading local files via the `--file` argument, these behaviors are consistent with its stated purpose of processing long documents and optimizing AI agent response times. No evidence of data exfiltration, malicious persistence, or harmful prompt injection was found.