troubleshooter

name	troubleshooter
description	Diagnoses and resolves errors when using Pletor MCP operations. Activate when any operation returns an error, when you've retried something more than once, or when you're stuck.

Troubleshooter

You are diagnosing an error with a Pletor MCP operation. Follow the guidance below to resolve it quickly.

When to Use This Skill

Activate when:

• Any Pletor MCP operation returns an error
• You've retried an operation more than once without success
• You're unsure how to proceed after a failure

Upload Failures

Symptom	Cause	Fix
file_path not found / ENOENT	Web MCP clients can't access local files	Use content_base64 instead. Read the file, base64-encode it, and pass with mime_type and filename.
"Could not determine MIME type"	No Content-Type header or unknown file extension	Pass mime_type explicitly (e.g., image/jpeg, video/mp4).
Upload succeeds but asset not usable	Forgot to provide filename	Always pass filename when using content_base64.

Authentication and Connection Errors

Symptom	Fix
search_workflows or other Pletor tool not found	Pletor MCP server isn't connected. Tell the user to run /mcp and enable Pletor. If not listed, they need to install the plugin first.
MCP call routed to wrong server (not Pletor)	Wrong MCP server is active. Tell the user to run /mcp, disable conflicting servers, and make sure Pletor is selected.
401 from any operation	MCP token expired. Tell the user to run /mcp, select Pletor, and re-authenticate.
"fetch failed" or network error	Backend API not reachable. Ask the user to check if Pletor is running.

Workflow Input Format Errors

This is the most common source of wasted retries. Never guess input formats.

The Golden Rule

Before passing inputs to run_workflow, always discover the schema:

1. Call get_workflow to read the DSL and identify input nodes (types ending with _input).
2. For well-known types, use these formats:
   • text_input → plain string
   • image_input / video_input / audio_input → { "asset_ids": ["<asset_id>"] }
3. For anything else → call get_node_details with the node type to get the expected schema.

When a Run Fails with Input Errors

Errors like "Field required", "Input should be a valid dictionary", or validation errors:

1. Call get_node_run_result for the failed node to read the actual error.
2. Call get_node_details for that node type to see the correct schema.
3. Compare what you sent vs what's expected. Fix and retry.
4. If it still fails after 2 attempts, the input may be pre-configured in the workflow. Try running without passing it in flow_input.

General Debugging Checklist

• Never retry the same input format twice. If it failed once, it will fail again.
• Always inspect errors before retrying. Call get_node_run_result to read the actual error message.
• Call get_node_details for unfamiliar node types. Don't guess schemas.
• Pre-configured inputs exist. If a workflow has been set up with default values for an input, you don't need to pass it. When stuck, try omitting the problematic input.
• Check the operation schema. Call get_schema with the operation name to verify you're passing the right parameters.