← Back to Skills Marketplace
248
Downloads
0
Stars
0
Active Installs
12
Versions
Install in OpenClaw
/install wucai
Description
五彩 (WuCai) 是 AI Agent 的个人知识库与第二大脑。它允许 AI 检索、总结和管理用户在网页上记录的划线、批注、全文剪藏及个人日记。 核心能力：跨区域语义搜索、时间区间知识追溯 (range lookup)、以及基于划线的精准内容问答。 它是 AI 了解用户兴趣偏好、进行每日知识复盘、以及在网页上...
Usage Guidance
This skill appears to do what it says: it wraps WuCai API calls in a local Python script and needs your WUCAI_API_TOKEN (and optionally WUCAI_REGION). Before installing: (1) Confirm how your OpenClaw platform stores credentials — do NOT paste your token into a chat if the platform will store it in plain conversation history; prefer using the platform's secure secret/credential input if available. (2) Verify you have Python (python3) available since the script requires it. (3) Note the regional data residency — pick the correct region (cn/eu/us) or you will see no data. (4) Understand the 14-day query span limit described in SKILL.md. If you need clarification on how the token will be stored by the platform or want the skill to accept tokens via a secure config UI instead of chat, ask the skill author or platform support before granting the credential.
Capability Analysis
Type: OpenClaw Skill
Name: wucai
Version: 26.3.37
The 'wucai' skill bundle is a legitimate integration for the WuCai Highlight service, designed to manage web annotations and personal notes. The core logic in `scripts/wucai_api.py` uses Python's native `urllib` to interact with official API endpoints (wucai.site, wucainote.com) and correctly handles authentication via the `WUCAI_API_TOKEN` environment variable. The instructions in `SKILL.md` are well-structured, emphasizing a 'script proxy' pattern to prevent shell injection and ensuring data isolation across regions (CN/EU/US) without any evidence of malicious intent or data exfiltration.
Capability Assessment
Purpose & Capability
Name/description, declared primary credential (WUCAI_API_TOKEN), provided Python wrapper (scripts/wucai_api.py) and the documented endpoints all match the stated purpose of accessing a user's WuCai highlights, notes and diary. Minor inconsistency: top-level registry metadata in the manifest summary listed 'no required binaries', while SKILL.md declares python/python3 in requires.commands — the script requires Python to run, so the skill implicitly needs Python on PATH.
Instruction Scope
SKILL.md tightly constrains how the agent must call the API (only via scripts/wucai_api.py using stdin) and the script only reads WUCAI_API_TOKEN and WUCAI_REGION — that's appropriate. However, the README explicitly instructs users to 'send the copied Token to the AI to bind', which encourages users to paste a secret into chat. That instruction risks exposing the token in conversation history or logs if the platform does not provide a secure secret storage workflow. The skill otherwise does not request or instruct reading unrelated files or credentials.
Install Mechanism
No install spec / remote downloads; all code is included in the skill bundle and the Python script uses only stdlib urllib. This is low-risk from an installer perspective — no external archives, no third-party package pulls.
Credentials
The only secret requested is WUCAI_API_TOKEN (and optional WUCAI_REGION) which is proportional to the described API integration. The script validates the token prefix and uses region-specific domains. The main concern is procedural: the documentation encourages users to paste the token into chat for 'binding' — that is a distribution method for a secret and not a technical requirement of the API itself.
Persistence & Privilege
always is false and the skill does not request persistent system-wide privileges or modify other skills. It runs on-demand via a local script which only issues outbound requests to the documented WuCai regional endpoints.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install wucai - After installation, invoke the skill by name or use
/wucai - Provide required inputs per the skill's parameter spec and get structured output
Version History
v26.3.37
- No functional changes or code modifications in this release.
- Documentation and specification remain consistent with the previous version.
- No updates to commands, API structure, or environment settings detected.
v26.3.36
- Updated SKILL.md and bundle contents.
v26.3.35
WuCai 26.3.35 Changelog
- Enhanced region separation: clarified that all data is physically isolated by region (cn, eu, us), and region changes connect to distinct databases.
- API interaction streamlined: removed the requirement for the auxiliary "read_clipped.py" script—now all API calls use only "wucai_api.py".
- Refined error handling and user guidance: error messages improved for cross-region scenarios, with clearer reference to region-specific records.
- Updated function scope: removed "read_clipped_content" support; summarization and analysis should use highlights and notes.
- Documentation and invocation examples revised for clarity and accuracy.
v26.3.34
- Removed scripts/read_clipped.py; now only scripts/wucai_api.py is required for API interaction.
- Updated range parameter logic: all time-based queries are now strictly limited to a maximum span of 14 days.
- Clarified error handling with explicit, bilingual user guidance for all major API error codes.
- Refined invocation protocol section to simplify required steps and improve clarity.
- Enhanced description to emphasize WuCai’s capabilities as a personal knowledge base and assistant for context-aware Q&A, knowledge review, and content management.
- Updated function documentation for stricter usage guidelines and improved parameter handling.
v26.3.33
- Added with_highlights parameter to list_articles and list_diary functions for direct inclusion of highlight/note content.
- Updated function documentation to clarify with_highlights option and usage.
- No breaking changes; API structure remains compatible.
v26.3.30
- Enforced a strict “script proxy only” protocol: All API interactions must use provided Python scripts; direct network requests are forbidden.
- Improved environment checks: Now explicitly requires both Python scripts (`wucai_api.py` and `read_clipped.py`) and compatible Python commands.
- Clarified invocation rules: Commands must use `echo ... | python3 ...` with JSON parameters via stdin.
- Enhanced region handling: AI responses must use Chinese for `cn` region, English for `eu/us`.
- Tightened guidance and error handling: OpenAPI links and help instructions must strictly match the user’s selected region.
v26.3.29
Major update: introduces Python脚本执行和全文剪藏读取，强化多区域支持与异常引导。
- 新增本地 Python 脚本（scripts/wucai_api.py, scripts/read_clipped.py）统一所有 API 调用，支持环境变量自动切换及15秒超时保护。
- 增加 read_clipped_content 功能，可主动读取剪藏文章的 Markdown 全文，用于摘要或深度问答。
- 强化动态区域引导和链接一致性，严格根据所选区域输出对应 URL。
- 明确异常处理规则（如Token缺失、权限受限）及反馈机制。
- 更新函数调用规范，更安全地传参和解析响应。
v26.3.28
Version 26.3.28
- 增加“动态引导约束”，要求根据当前 $WUCAI_REGION 区域严格输出匹配链接。
- 明确规定输出“获取 TOKEN”及“会员中心”链接时，必须参考 api-details.md 的区域信息表，确保链接一致性。
- 其余 API 函数与逻辑保持不变。
v26.3.27
- Improved parameter descriptions and usage examples in function definitions for clarity.
- Clarified optional values for the status parameter in diary listing.
- Updated terminology for consistency (e.g., "将文章移动到回收站" instead of "逻辑删除文章").
- Removed mention of a "read_clipped_content" function and related scripting details.
- Enhanced overall accuracy and brevity in both function and core rules descriptions.
v26.3.26
WuCai skill now supports multi-region (CN/EU/US) selection and bilingual interaction.
- 新增环境变量 WUCAI_REGION，实现跨区域 (CN/EU/US) API 支持。
- 支持中英文双语自动识别与交互，自动切换核心术语。
- 大幅扩展 API 调用：文章、日记、划线的分页检索/搜索/详情/批注等功能。
- 支持全文剪藏与日记追加。
- 自动处理分页、异常与权限提示。
- 配置文档和接口调用规则全面更新。
v26.3.25
wucai v26.3.25
- Added _meta.json file.
- Updated API base URL from banana.wucai.site to api.wucai.site.
- Clarified API token format to use "wct-" prefix in Authorization header.
- Minor formatting and structure improvements in documentation.
v26.3.24
- Initial release of "五彩": a web highlighting, annotation, note-taking, and personal knowledge management skill.
- Added quick-save commands for creating notes from any text using trigger words like “记一下”, “存到笔记”, or “帮我记录”.
- Provided onboarding instructions for authorization using user-generated API TOKEN.
- Enforced privacy by not displaying user note contents proactively in group chats.
- Integrated error code handling, membership validation, and API usage recommendations.
- Included clear user guidance for connecting and configuring the skill.
Metadata
Frequently Asked Questions
What is WuCai?
五彩 (WuCai) 是 AI Agent 的个人知识库与第二大脑。它允许 AI 检索、总结和管理用户在网页上记录的划线、批注、全文剪藏及个人日记。 核心能力：跨区域语义搜索、时间区间知识追溯 (range lookup)、以及基于划线的精准内容问答。 它是 AI 了解用户兴趣偏好、进行每日知识复盘、以及在网页上... It is an AI Agent Skill for Claude Code / OpenClaw, with 248 downloads so far.
How do I install WuCai?
Run "/install wucai" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is WuCai free?
Yes, WuCai is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does WuCai support?
WuCai is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created WuCai?
It is built and maintained by makediff (@makediff); the current version is v26.3.37.
More Skills