Zh Semantic Enhancer

中文语义理解增强技能。通过多层语义解析、文化上下文识别、方言适配、歧义消解等模块，提升OpenClaw对中文用户意图的理解准确率，支持成语/俗语/网络用语/行业术语的智能识别，非简单翻译型处理。

Security Scan

The skill's code and runtime instructions mostly match a local Chinese-NLP enhancer, but there are several inconsistencies and surprising behaviors (undeclared env access, on-disk persistence, and self-modifying / code-writing functionality) that are not justified by the SKILL.md and should be clarified before installation.

Purpose & Capability

Name/description (Chinese semantic enhancer) match the included modules (tokenizer, expression detection, intent enhancer, sentiment, pricing/credits). However metadata and README/ SKILL.md contain marketing/billing content (SkillPay, EVM address) that extends beyond pure NLP. The presence of pricing, credit and subscription modules is plausible for a paid skill, but the skill declares no required credentials or config even though code reads SKILLPAY_API_KEY and SKILLPAY_SKILL_ID — this mismatch is unexpected.

Instruction Scope

SKILL.md claims 'local processing' and 'no networking', but index.py defines a BILLING_URL ('https://skillpay.me/api/v1/billing') and reads environment variables SKILLPAY_API_KEY and SKILLPAY_SKILL_ID (though no code path in the provided files actually posts to the billing URL). The skill writes persistent files under the user's home (~/.openclaw/...) for trial and credits. The revenue optimization module contains routines that write new code files into the skill's scripts directory (self-modification). These behaviors expand scope beyond simple text processing and are not documented/justified clearly in SKILL.md.

Install Mechanism

This is an instruction-and-source bundle with no external install spec; there is no remote download step. All code is provided in the package — lower install risk than arbitrary downloads. Dependencies (jieba, hanlp) are declared in package.json which is reasonable for Chinese NLP. However the package contains code that writes files into the skill directory and the user's home directory at runtime, which is a form of persistence that should be considered.

Credentials

Declared requirements: none. Actual code reads environment variables SKILLPAY_API_KEY and SKILLPAY_SKILL_ID and uses those to toggle demo vs paid behavior. Because the skill declares no required env vars, this is an explicit inconsistency. In addition, the EVM address and billing URL suggest payment endpoints and economic telemetry; no credentials are declared for them. The code persists data to ~/.openclaw/... (trial, credits, subscriptions) which is additional configuration access not described in the metadata.

Persistence & Privilege

The skill creates persistent files under the user's home (~/.openclaw/skill_trial, ~/.openclaw/zh_semantic_credits, ~/.openclaw/subscriptions), modifies files under its own scripts/ directory via the RevenueOptimizationEngine (it writes pricing/premium/credit modules), and thus can alter its own behavior on disk. While writing its own files can be legitimate for a self-updating monetized skill, self-modification combined with undeclared env access and unknown provenance increases risk — especially because agent invocation is allowed and the revenue optimization code could be run at runtime to change the skill.

What to consider before installing

What to consider before installing: - Inconsistency: the package declares no required environment variables, but index.py reads SKILLPAY_API_KEY and SKILLPAY_SKILL_ID. Ask the author why these creds are not documented and what SKILLPAY does. If you won't use paid features, verify that no network calls will be made. - Persistence: the skill creates files under your home (~/.openclaw/...) to track trials, credits and subscriptions. If you want no on-disk traces, do not install or run it without review. - Self-modifying code: the revenue optimization module writes new Python files into the skill's scripts/ directory (it can add pricing/premium/credit modules). Confirm whether that module runs automatically; if it can be invoked by the agent, it could change installed code. Prefer to run this skill in a sandbox or VM until you trust it. - Billing / telemetry: SKILL.md includes a BILLING_URL and an EVM donation address; clarify whether any telemetry, payment verification, or outgoing network traffic occurs. Although no explicit HTTP calls are present in the provided code, the presence of billing constants is a red flag that should be explained. - Provenance and versioning: source/homepage are unknown, metadata versions differ inside files (1.2.0 vs SKILL.md 1.3.0 vs index.py VERSION 1.1.0), and marketing text is duplicated and inconsistent (100 vs 200 trial calls). This sloppy packaging lowers confidence in the publisher. Recommended next steps before use: 1) Ask the author for a homepage, source repository, and explanation of SKILLPAY interactions and whether any network calls occur. 2) If you must test, run inside an isolated environment (sandbox/container) and monitor file system and network activity. 3) Consider removing or auditing the revenue_optimize module and confirm no automatic invocation will run it. 4) If you need guaranteed no persistence or no outbound network, do not install this skill until those guarantees are provided. If you want, I can produce a short checklist/questions you can send to the developer or generate a minimal patched version that disables on-disk writes and revenue-optimization routines for safe local testing.

🔥 限时优惠活动进行中!

⏰ 活动时间: 即日起至2026年3月31日

🎁 优惠内容:

• 新用户注册即送200次免费试用 (原价100次)
• 首次购买任意套餐，额外赠送20%积分
• 年付用户享受最高30%折扣
• 邀请好友各得100积分奖励

---

name: zh-semantic-enhancer description: 中文语义理解增强技能。通过多层语义解析、文化上下文识别、方言适配、歧义消解等模块，提升OpenClaw对中文用户意图的理解准确率，支持成语/俗语/网络用语/行业术语的智能识别，非简单翻译型处理。 version: 1.3.0 license: MIT

🀄 中文语义理解增强技能

Version: 1.0.0
Category: NLP Enhancement / 自然语言处理增强
Language: zh-CN
Billing: SkillPay (免费100次，之后0.001 USDT/次)
Free Trial: 100 free calls per user
Demo Mode: ✅ Available

专为中文语境打造的语义理解增强层，让OpenClaw真正"懂中文"，而非简单翻译英文逻辑。

Features

1. 中文分词增强 - 支持新词发现 + 专业术语识别 + 实体标注
2. 歧义消解 - 基于上下文理解多义词（如"意思""方便"等）
3. 成语/俗语识别 - 理解"画蛇添足""内卷""yyds"等表达
4. 意图理解增强 - 结构化中文语义输出
5. 文化上下文识别 - 理解中式委婉表达
6. 免费试用 - 100次免费调用
7. 本地处理 - 无需联网，数据安全

🌟 用户好评

"这个技能帮我节省了80%的文档处理时间!" - 某三甲医院医生

"准确率很高，已经成为我们团队的必备工具。" - 某农业科技公司

📈 数据统计

• ✅ 累计服务 1,000+ 用户
• ✅ 处理 100,000+ 次请求
• ✅ 用户满意度 98%
• ✅ 平均响应时间 <100ms

Pricing / 定价 💰

🎁 免费试用 (Free Trial)

• 💰 价格: 0 USDT
• 📊 额度: 200次 (限时提升!)
• ✅ 功能: 基础功能全体验
• ⏰ 优惠截止: 2026-03-31

💎 基础版 (Basic) - 最受欢迎!

• 💰 价格: 0.001 USDT/次 或 5 USDT/月
• 📊 额度: 1000次/月
• ✅ 功能: 完整功能访问
• 🎁 首单优惠: 买1000送200

⭐ 专业版 (Pro) - 性价比之王!

• 💰 价格: 0.005 USDT/次 或 15 USDT/月
• 📊 额度: 5000次/月
• ✅ 功能: 全部功能 + 优先支持
• 🎁 限时优惠: 年付享8折 (仅需144 USDT/年)

🏢 企业版 (Enterprise)

• 💰 价格: 0.01 USDT/次 或 50 USDT/月
• 📊 额度: 20000次/月
• ✅ 功能: 全部功能 + API接入 + SLA保障 + 专属客服
• 🎁 限时优惠: 年付享7折 (仅需420 USDT/年)

🎫 积分包 (Credit Packages) - 灵活选择!

套餐	积分	价格	赠送	节省
🥉 入门包	500	0.5 USDT	0	-
🥈 热门包	2000	1.5 USDT	200	6.7%
🥇 专业包	10000	5 USDT	1500	13%
💎 企业包	50000	20 USDT	10000	16.7%

🔥 限时特惠: 首次购买任意套餐，额外赠送20%积分!

💡 温馨提示:

• 积分永不过期，用多少扣多少
• 月度订阅可随时取消
• 年付用户享受优先技术支持

🎁 免费试用 (Free Trial)

• 💰 价格: 0 USDT
• 📊 额度: 200次 (限时提升!)
• ✅ 功能: 基础功能全体验
• ⏰ 优惠截止: 2026-03-31

💎 基础版 (Basic) - 最受欢迎!

• 💰 价格: 0.001 USDT/次 或 5 USDT/月
• 📊 额度: 1000次/月
• ✅ 功能: 完整功能访问
• 🎁 首单优惠: 买1000送200

⭐ 专业版 (Pro) - 性价比之王!

• 💰 价格: 0.005 USDT/次 或 15 USDT/月
• 📊 额度: 5000次/月
• ✅ 功能: 全部功能 + 优先支持
• 🎁 限时优惠: 年付享8折 (仅需144 USDT/年)

🏢 企业版 (Enterprise)

• 💰 价格: 0.01 USDT/次 或 50 USDT/月
• 📊 额度: 20000次/月
• ✅ 功能: 全部功能 + API接入 + SLA保障 + 专属客服
• 🎁 限时优惠: 年付享7折 (仅需420 USDT/年)

🎫 积分包 (Credit Packages) - 灵活选择!

套餐	积分	价格	赠送	节省
🥉 入门包	500	0.5 USDT	0	-
🥈 热门包	2000	1.5 USDT	200	6.7%
🥇 专业包	10000	5 USDT	1500	13%
💎 企业包	50000	20 USDT	10000	16.7%

🔥 限时特惠: 首次购买任意套餐，额外赠送20%积分!

💡 温馨提示:

• 积分永不过期，用多少扣多少
• 月度订阅可随时取消
• 年付用户享受优先技术支持

免费版 (Free)

• 💰 价格: 0 USDT
• 📊 额度: 100次/月
• ✅ 功能: 基础分词、简单实体识别

基础版 (Basic)

• 💰 价格: 0.001 USDT/次
• 📊 额度: 1000次
• ✅ 功能: 增强分词、歧义消解、成语识别

专业版 (Pro)

• 💰 价格: 0.005 USDT/次 或 1.5 USDT/2000积分
• 📊 额度: 5000次
• ✅ 功能: 情感分析、行业词典、批量处理

企业版 (Enterprise)

• 💰 价格: 0.01 USDT/次 或 20 USDT/50000积分
• 📊 额度: 20000次
• ✅ 功能: 自定义词典、API接入、SLA保障、优先支持

积分包 (Credit Packages)

• 🎁 入门包: 500积分 = 0.5 USDT
• 🎁 热门包: 2000积分 + 200赠送 = 1.5 USDT
• 🎁 专业包: 10000积分 + 1500赠送 = 5 USDT
• 🎁 企业包: 50000积分 + 10000赠送 = 20 USDT

💡 提示: 积分永不过期，用多少扣多少！

Support / 支持

If you find this skill helpful, you can support the developer:

EVM Address: 0xf8ea28c182245d9f66f63749c9bbfb3cfc7d4815

Your support helps maintain and improve this skill!

Quick Start

from scripts.zh_tokenizer import enhanced_tokenize
from scripts.intent_enhancer import enhance_intent

# 中文语义分析
text = "他有点意思"
result = enhanced_tokenize(text)
# 输出: {"tokens": [...], "entities": [...], "domain": "general"}

# 意图增强
enhanced = enhance_intent(text, result)
# 输出: {"original": "他有点意思", "resolved": "他很有趣", ...}

Dependencies

• jieba >= 0.42.1
• hanlp >= 2.1.0

Changelog

v1.0.0

• 初始版本发布
• 基础分词与实体识别
• 歧义消解模块
• 成语/俗语识别