Wegent Knowledge

What the source does

• FilesystemNone observedNo reads or writes in the source
• NetworkNone observedNo outbound calls in the source
• ShellNone observedNo commands run in the source
• DependenciesPinnedNothing floats
• SecretsNone foundNo keys in source

Five audit layers

Gates applied: no behavioural pass.

backend-rs/config/example.env

CLAUDE.md

backend-rs/CLAUDE.md

Audit 0.4.1 · source sha 52a5031418c3 · 16 Oct 2026Report an issue or request a re-scan

• Publisherwecode-ai GitHub organizationOwner of the repository this skill is read from
• Source folderwecode-ai/Wegent/backend/init_data/skills/wegent-knowledgeThe exact folder we audited
• LicenceApache-2.0As the repository states it
• Commit pinned for this verdict52a5031418c3A new audit runs when the source changes
• Repository statusNot archivedPushed today
• Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

Its description: Knowledge base management and search tools for Wegent. Provides capabilities to list, create, update, and search knowledge bases and documents using RAG retrieval. Use this skill when the user wants to manage knowledge…

---
description: "Knowledge base management and search tools for Wegent. Provides capabilities to list, create, update, and search knowledge bases and documents using RAG retrieval. Use this skill when the user wants to manage knowledge bases, documents, or search for information programmatically."
displayName: "知识库工具"
version: "1.1.1"
author: "Wegent Team"
tags: ["knowledge", "knowledge-base", "document", "rag"]
bindShells:
  - Chat
  - Agno
  - ClaudeCode
mcpServers:
  wegent-knowledge:
    type: streamable-http
    # NOTE: MCP client only supports ${{...}} variable substitution.
    # The platform will inject `backend_url` via task_data.
    url: "${{backend_url}}/mcp/knowledge/sse"
    headers:
      Authorization: "Bearer ${{task_token}}"
    timeout: 300
---

# Wegent Knowledge Base Skill

You now have access to Wegent Knowledge Base management tools.

## Available Tools

- **wegent_kb_list_knowledge_bases**: List all knowledge bases accessible to the current user
  - scope: "personal" (your own), "group" (team), "organization", or "all" (default)
  - group_name: Specify group name when scope="group"
  - limit: Maximum number of knowledge bases to return per page
  - offset: Start offset for paginated listing
  - returns: items, total, returned_count, limit, offset, has_more
  - each item includes namespace_level ("personal", "group", or "organization") and namespace_display_name ("personal", group display name, or organization display name)

- **wegent_kb_list_documents**: List all documents in a knowledge base
  - knowledge_base_id: ID of the knowledge base
  - folder_id: Optional selected folder ID
  - include_subfolders: Whether to include descendant folders (default: true)
  - limit: Maximum number of documents to return per page
  - offset: Start offset for paginated listing
  - returns: items, total, returned_count, limit, offset, has_more

- **wegent_kb_create_knowledge_base**: Create a new knowledge base
  - name: Knowledge base display name
  - description: Optional description
  - namespace: "default" (personal) or a group namespace (requires Maintainer+)
  - kb_type: "notebook" (default) or "classic"
  - summary_enabled: Enable automatic summary generation (default: true)

- **wegent_kb_create_document**: Create a new document in a knowledge base
  - knowledge_base_id: Target knowledge base ID
  - name: Document name
  - source_type: "text" (paste content), "file" (base64 encoded), or "web" (URL to scrape)
  - content: Document content when source_type="text"
  - file_base64: Base64 encoded file when source_type="file"
  - file_extension: File extension when source_type="file"
  - url: URL to fetch when source_type="web"
  - trigger_indexing: Whether to trigger RAG indexing (default: true)
  - trigger_summary: Whether to trigger summary generation (default: true)

- **wegent_kb_read_document_content**: Read raw document content with offset/limit pagination
  - document_id: Document ID to read
  - offset: Character offset to start reading from (default: 0)
  - limit: Maximum number of characters to return (uses the backend default when omitted)
  - returns: content slice, total_length, returned_length, has_more, kb_id

- **wegent_kb_update_document_content**: Update a document's content for text documents and editable plain-text files
  - document_id: Document ID to update
  - content: New content (replaces existing content)
  - trigger_reindex: Whether to trigger RAG re-indexing (default: true)

- **wegent_kb_search_knowledge_base**: Search documents using RAG retrieval
  - knowledge_base_id: Knowledge base ID to search
  - query: Search query text
  - search_hints: Optional query plan. Only use `semantic_query` (string), `keywords` (array of terms), and `phrases` (array of exact multi-word matches). Never use `exact_keywords`.
  - max_results: Maximum results to return (default: 10, max: 50)
  - document_ids: Optional list of document IDs to filter search scope
  - folder_ids: Optional list of selected folder IDs
  - include_subfolders: Whether folder IDs include descendants

## Usage Notes

- All operations inherit the current user's permissions
- For a folder in `<selected_knowledge_sources>`, pass its original ID through `folder_id`/`folder_ids` and set `include_subfolders=true`. A selected folder always includes its entire descendant subtree. Never broaden the request to the whole knowledge base to discover descendants.
- For a selected document, pass its original ID through `document_ids` instead of broadening to the whole knowledge base.
- After creating or updating documents, indexing happens asynchronously
- Documents may show status "pending" until indexing completes
- For web scraping, the URL content is fetched and stored as document content
- `wegent_kb_update_document_content` supports `text` documents and plain-text file documents such as `txt`, `md`, and `markdown`; binary files like `pdf` or `docx` still require creating or replacing the source file instead of inline editing
- Avoid loops: if a tool call fails, report the error once and stop retrying/re-loading the skill unless the user changes inputs.
- Long documents should be read incrementally: start with the backend default limit, then continue with `offset = previous_offset + previous_returned_length` while `has_more=true`

## Example Workflow

1. List documents in a specific knowledge base:
   ```
   wegent_kb_list_documents(knowledge_base_id=123, limit=20, offset=0)
   ```

   If `has_more=true`, continue with:
   ```
   wegent_kb_list_documents(knowledge_base_id=123, limit=20, offset=20)
   ```

2. Create a new knowledge base:
   ```
   wegent_kb_create_knowledge_base(
     name="My KB",
     description="My personal notes",
     namespace="default",
     kb_type="notebook"
   )
   ```

3. Create a new text document:
   ```
   wegent_kb_create_document(
     knowledge_base_id=123,
     name="Meeting Notes",
     source_type="text",
     content="Notes from today's meeting..."
   )
   ```

4. Up

Excerpt, first 6,000 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

git clone https://github.com/wecode-ai/Wegent
git -C Wegent checkout 52a5031418c3

Works in

1. 2026-10-16
   ✓SAFE B 89 / 100

   first audit · source 52a5031418c3

Questions about Wegent Knowledge

• github.com/wecode-ai/Wegent/tree/main/backend/init_data/skills/w — the audited source at commit 52a5031418c3 and every finding aboveObserved · 16 Oct 2026

Spotted something wrong? Report a correction · Methodology