| name | comfyui-api-workflow-automation |
| description | Workflow for integrating new ComfyUI workflows into an MCP server and handling directory restriction errors. |
ComfyUI API Workflow Automation (Symlink vs Copy)
This skill provides a workflow for integrating new or shared ComfyUI workflows into an active MCP server environment.
Trigger Conditions
- When adding new
.json workflows to a remote ComfyUI server via MCP.
- When encountering "Attempted to read workflow outside of the configured directory" errors despite files being visible in
list_workflows.
Workflow Steps
- Identify Source: Locate the target workflow (e.g., in a Google Drive shared folder).
- Initial Integration (Symlink): Attempt to symlink the file to the ComfyUI root directory using
ln -sf <source> <target>. This is preferred for real-time updates.
- Verification: Run
mcp_comfyui_mcp_list_workflows to ensure the file is visible.
- Test Read: Attempt
mcp_comfyui_mcp_read_workflow on the new file.
- Error Handling (The "Symlink Trap"):
- If
read_workflow returns an error stating the file is "outside of the configured directory" even though it appears in the list, the MCP server is failing to resolve the symlink.
- Resolution (Hard Copy):
- Instead of a symlink, perform a direct copy:
cp <source> <target>.
- Final Verification: Re-run
mcp_comfyui_mcp_read_workflow to confirm successful access.
Pitfalls & Lessons Learned
- Symlink Resolution: Some MCP server implementations do not follow symlinks for security reasons, even if the link points to a valid path within the allowed root. A hard copy is the most reliable fallback.
- Sandbox Limitations: Do NOT attempt to use
execute_code (Python) to call MCP tools directly; they are agent-level tools and cannot be imported into the Python sandbox. Use direct tool calls instead.
Node Mapping Standard
When automating via API, always inspect the JSON structure first to map:
- Prompt Text Node ID (e.g.,
6)
- Resolution/Latent Node ID (e.g.,
8)
