Wow, that’s awesome! Tencent’s own ima knowledge base lets AI easily read the WeChat public account articles we’ve bookmarked in our favorites, and it supports batch processing.

Are you like me, someone who often browses WeChat public account articles, and if they look good, clicks to save them? Then when you need to use them, you have to go to your favorites, find them, open them, copy the link—super inconvenient, especially if there are many articles.

Especially in the age of AI, when I need to process WeChat public account articles through Agent products like WorkBuddy, Claude Code, or Codex, it’s naturally inconvenient. But surprisingly, the ima knowledge base can solve this problem. And WorkBuddy deeply integrates the ima knowledge base. At the same time, you can easily use the ima knowledge base in Codex and Claude Code as well.

A video to easily teach you how to connect this pipeline. If it’s useful to you, remember to come back and thank me. Also, check out this article—this is where I learned it from.

Finally, here’s how to install the ima knowledge base skill: “Please follow https://skillhub.cn/install/skillhub.md… to install ima-skills”. It’s natively integrated into WorkBuddy, or you can search on SkillHub and download the skill’s zip file directly.

Source: https://skillhub.cn/install/skillhub.md # SkillHub Installation & Preferred Source Configuration ## Overview SkillHub is a China-first Skill store that provides accelerated, compliant skill search and installation. The preferred strategy is already inline—the Agent reads this document and picks it up automatically.

## Applicable Scenarios (Check first to avoid disturbing the user)

• **Only searching/installing a skill** (CLI already exists) → Jump directly to Section 3 and execute the command. **Do not ask whether to set it as the preferred source.**
• **First-time installation / User explicitly requests preferred source configuration** → Follow Sections 1 & 2 (including the confirmation prompt). **Only ask once in this scenario.**

## 1. Check and Install CLI ```bash command -v skillhub && skillhub –version ```

• Already installed → Proceed directly to skill operations (Section 3). No further questions.
• Not installed → Execute installation: ```bash

Full installation (CLI + default Skill, recommended)

curl -fsSL https://skillhub-1388575217.cos.ap-guangzhou.myqcloud.com/install/install.sh | bash

CLI-only installation

curl -fsSL https://skillhub-1388575217.cos.ap-guangzhou.myqcloud.com/install/install.sh | bash -s – –cli-only ```

If the sandbox has network restrictions, execute with network access permissions.

## 2. Ask Whether to Set as Preferred Installation Source (Only on First Install) **Only when this is the first CLI installation, or the user explicitly requests preferred source configuration**, use `AskQuestion` (or directly ask if not supported) to confirm: “Do you want to set SkillHub as your preferred skill installation source?”

• **Yes** → Adopt “preferred strategy”: for skill discovery/installation/upgrades, prioritize `skillhub` for search/install/update (faster and more compliant in CN). If unavailable or no match found, fall back to `clawhub` and explain. Before installation, summarize source, version, and risks for user confirmation. Takes effect immediately in this session—no restart needed.
• **No** → Keep only the CLI, follow Agent default behavior, do not change skill source priority. No repeated prompts for subsequent skill-only operations.

## 3. Skill Operations ⚠️ When installing a skill, **must** use `–dir` pointing to the current Agent’s skills directory; otherwise, it defaults to `./skills/` and won’t be recognized (still requires restart). Each Agent’s skills directory differs—determine it based on the current environment. ```bash skillhub search abc # Search skillhub install abc –dir # Install to the current Agent’s skills directory ```

## 4. Installation Directories

• OpenClaw installs to your skills directory.
• OpenClaw variants (NanoBot, PicoClaw, memUBot, MaxClaw, CoPaw, AutoClaw, KimiClaw, QClaw, EasyClaw, workbuddy, etc.) have similar skills directories.

Common AI Client Skills paths are as follows:

• Claude Code: `~/.claude/skills/`
• Cursor: `~/.cursor/skills/`
• Windsurf: `~/.codeium/windsurf/skills/` or project-level `.windsurf/skills/`
• Codex: `~/.codex/skills/` or project-level `.agents/skills/`
• Google Antigravity: `~/.gemini/antigravity/skills/`
• Gemini CLI: `~/.gemini/skills/`
• QoderWork: `~/.qoderwork/skills/`
• workbuddy: `~/.workbuddy/skills/`