Modifies existing Corezoid processes and fixes broken connectors — covers exporting, editing process JSON, managing environment variables, and diagnosing failed tasks. Activate whenever a user asks to update a process, fix a connector, add a node, change how a process works, debug a failed task, or provides ResultNodeObjID and ResultTaskData showing where execution stopped. Also activate when the user says "add a call to process X", "the task failed with error Y", or "modify the logic". This skill knows the correct export→edit→variable→test workflow and error taxonomy.

Corezoid Process Updater

Name: corezoid-process-updater
Author: corezoid

CRITICAL: Use MCP tools for all API operations

For reading process state and modifying nodes, always use MCP tools — never write Python scripts:

Operation	Use this MCP tool
Read process structure	`mcp__corezoid__get_process_scheme`
Get process metadata	`mcp__corezoid__get_process_details`
Modify a single node	`mcp__corezoid__modify_node`
Modify multiple nodes	`mcp__corezoid__batch_modify_nodes`
Commit changes	`mcp__corezoid__commit`
Create a test task	`mcp__corezoid__create_task`
Check task result	`mcp__corezoid__show_task`
Trace task execution	`mcp__corezoid__list_task_history`
Any raw API op	`mcp__corezoid__make_request`

convctl.sh is still used for fetch/deploy operations (it's the CLI deployment tool, not a Python replacement).

Required parameters

Parameter	Purpose
`API_URL`	Corezoid instance URL
`API_TOKEN`	Auth token
`WORKSPACE_ID`	Workspace identifier
`PROJECT_ID`	Project identifier (needed for variable creation)
`ROOT_FOLDER_ID`	Root folder for export
`PROC_ID`	Process ID being modified

Path A — Updating existing logic

1. Export the project context

API_URL=<API_URL> API_TOKEN=<API_TOKEN> WORKSPACE_ID=<WORKSPACE_ID> \
  ./convctl.sh fetch-folder <ROOT_FOLDER_ID> .processes/

Analyze the exported processes to understand the current structure and any process IDs referenced.

2. Export the specific process to edit

API_URL=<API_URL> API_TOKEN=<API_TOKEN> WORKSPACE_ID=<WORKSPACE_ID> \
  ./convctl.sh fetch-process <PROC_ID> .processes/target_process.json

Work on .processes/target_process.json — this is the file to edit.

3. Make the changes

Apply the requested modifications directly to .processes/target_process.json.

If adding a Call Process node, use this structure:

{
  "id": "<new-node-id>",
  "obj_type": 0,
  "condition": {
    "logics": [
      {
        "type": "api_rpc",
        "conv_id": <target_process_id>,
        "err_node_id": "<reply-error-node-id>",
        "extra": {
          "param1": "{{value1}}",
          "param2": "{{value2}}"
        },
        "extra_type": {
          "param1": "string",
          "param2": "number"
        },
        "group": "all"
      },
      {"type": "go", "to_node_id": "<next-node-id>"}
    ],
    "semaphors": []
  },
  "title": "Call <ProcessName>",
  "description": "Calls the X process with parameters",
  "x": 500, "y": 460,
  "extra": "{\"modeForm\":\"expand\",\"icon\":\"\"}",
  "options": null
}

CRITICAL: type: "api_rpc" — never use "call_process". extra and extra_type must have identical keys with matching types.

4. Manage variables (for new constants)

All URLs, tokens, API endpoints, and configuration values must be stored as variables — never hardcoded.

Check first: look in .processes/variables.json — if the variable already exists, use it.

Create if missing:

API_URL=<API_URL> API_TOKEN=<API_TOKEN> WORKSPACE_ID=<WORKSPACE_ID> \
  PROJECT_ID=<PROJECT_ID> \
  ./convctl.sh create-variable <ROOT_FOLDER_ID> <NAME> <DESCRIPTION> <VALUE>

Variable naming rules: lowercase letters, numbers, hyphens only — min 3 chars. Good: api-endpoint, auth-token-prod, db-host. Bad: apiEndpoint, token, T.

Reference in process JSON: {{env_var[@variable-name]}}

5. Test

API_URL=<API_URL> API_TOKEN=<API_TOKEN> WORKSPACE_ID=<WORKSPACE_ID> \
  ./convctl.sh run-process <PROC_ID> .processes/target_process.json '{"key":"value"}'

Must pass ("Task completed") before the work is done.

Path B — Fixing a broken connector (error diagnosis)

1. Collect error information from the user

Ask for:

ResultNodeObjID — the node ID where execution stopped
ResultTaskData — the full task data object at the time of failure

2. Export the current process

API_URL=<API_URL> API_TOKEN=<API_TOKEN> WORKSPACE_ID=<WORKSPACE_ID> \
  ./convctl.sh fetch-process <PROC_ID> .processes/target_process.json

Find the node with "id": "<ResultNodeObjID>" in the JSON.

3. Diagnose the error

Read ResultTaskData and look for these system parameters:

For API Call failures (api logic):

Parameter	Meaning
`__conveyor_api_return_type_error__`	`"hardware"` = network/infra issue (retryable), `"software"` = logical error
`__conveyor_api_return_code__`	HTTP status code returned
`__conveyor_api_return_type_tag__`	Error tag: `api_connection_error`, `api_bad_answer`, `api_timeout`, `api_validation_error`
`__conveyor_api_return_description__`	Error description string

For Code Node failures (api_code logic):

Parameter	Meaning
`__conveyor_code_return_type_error__`	`"hardware"` or `"software"`
`__conveyor_code_return_description__`	JS error message (syntax error, runtime exception)
`__conveyor_code_return_type_tag__`	`api_code_syntax_error` or `api_code_runtime_error`

Common error patterns and fixes:

Error tag	Likely cause	Fix
`api_connection_error`	URL unreachable or DNS failure	Check URL variable, verify endpoint availability
`api_bad_answer` + `no_scheme`	URL missing `https://` prefix	Fix URL value in env_var or code node
`api_bad_answer` (4xx/5xx)	API returned an error	Check request body, headers, auth token
`api_timeout`	External API too slow	Increase semaphore timeout or check API
`api_code_syntax_error`	JS syntax error in Code node	Fix the `src` field — check for unclosed brackets, typos
`api_code_runtime_error`	JS runtime exception	Add null checks, verify data fields exist before use
`api_validation_error`	Request didn't match API schema	Check `raw_body` structure against API spec

4. Fix the failing node

Apply the fix to .processes/target_process.json and re-test:

API_URL=<API_URL> API_TOKEN=<API_TOKEN> WORKSPACE_ID=<WORKSPACE_ID> \
  ./convctl.sh run-process <PROC_ID> .processes/target_process.json '<input_task_data>'

Repeat until "Task completed".