Tencent Docs

The Problem to Solve

When building automation workflows or AI Agents, there's a frequent need to programmatically create, read, or edit cloud documents. However, different document types (e.g., PPT, Excel, Smart Canvas) have their own distinct APIs and internal structures. Developers must adapt to multiple API endpoints, handling varied parameters and return formats, which introduces significant integration complexity and maintenance overhead. A specific and common need is how to efficiently and securely operate on all file types within the Tencent Docs ecosystem using a unified approach.

How the Skill Works

The Tencent Docs MCP skill encapsulates this complexity into a structured set of tools and routing rules.

1. Document Type Routing & Tool Set Separation

The skill's core is the Scenario Routing Table. It first identifies the document category (e.g., sheet, slide, smartcanvas) via file ID or link prefix, then strictly routes to the corresponding tool set. For example:

• For PPT / Slides (file link contains /slide/ or when "presentation" is mentioned), it must use the slide-mcp service's slide_* tools and follow its dedicated workflow.
• For Smart Canvas, use create_smartcanvas_by_mdx for creation or smartcanvas.* tools for editing.
• For Excel, it uses the sheet-mcp's sheet.* tools.

This design ensures each document type is handled by APIs specifically designed for it, preventing structural corruption from using incorrect tools (corresponding to error code 400016).

2. Key Workflows

• Document Creation: Routes to corresponding create_* methods based on intent keywords (e.g., "report," "PPT," "mind map"). For instance, use create_mind_by_markdown for mind maps and create_flowchart_by_mermaid for flowcharts.
• Content Editing & Management: Editing requires first determining the original file type before using the appropriate tools. File management operations (rename, move, permissions) use the manage.* tool set.
• Batch Operations: To optimize performance and cost, when performing consecutive multiple writes to the same document (e.g., inserting multiple rows into a sheet), the skill mandates using batch interfaces (like smartsheet.add_records) for a single submission instead of looping through single-write interfaces.

3. Public Capabilities & Conversions

The skill also provides cross-type public capabilities:
- get_content: A universal interface for reading document content.
- Web Clipping: Converts URL content into a Smart Canvas document via scrape_url.
- Local File Upload: Uses the import_file.sh script to import local files.
- OCR Image Recognition: Converts images into Word or Excel documents.

Applicable Boundaries & Caveats

• Special Case for PPT Tasks: All PPT-related needs (generation, editing, continuation) must be handled through the slide-mcp's dedicated tool chain. Using doc-mcp or general tools will result in incorrect content or behavior.
• Authorization & Quotas: Local authorization must be completed before use (error code 400006). Lack of VIP permissions (400007) or credits (400008) will restrict certain functions.
• Parameters & Formats: The mdx parameter in document creation tools directly supports Markdown syntax. For OCR, you must choose between a public URL or base64; local images must be processed via a specific script.
• Asynchronous Tasks: For asynchronous operations like web clipping, it's recommended to use a dedicated sub-session for polling to avoid blocking the main interaction flow.

Essentially, this skill serves as a protocol layer for developers. It aggregates disparate document APIs logically and enforces best practices (like type matching and batch writes), making it possible to build reliable document automation workflows.