# diagnostics

> List the current running model, capabilities, and validate tools/skills. There is no function named 'diagnostics'; use the exec tool to run: node /app/scripts/diagnostics/cli.js --section <section>.

- Author: Corey Szopinski
- Repository: coreindustries/core-openclaw-docker
- Version: 20260201075255
- Stars: 2
- Forks: 0
- Last Updated: 2026-02-06
- Source: https://github.com/coreindustries/core-openclaw-docker
- Web: https://mule.run/skillshub/@@coreindustries/core-openclaw-docker~diagnostics:20260201075255

---

---
name: diagnostics
description: "List the current running model, capabilities, and validate tools/skills. There is no function named 'diagnostics'; use the exec tool to run: node /app/scripts/diagnostics/cli.js --section <section>."
metadata: {}
---

## Triggers

This skill activates when a user message matches any of the following:

- `/diagnostics`
- "what model am I using" / "which model is running"
- "show diagnostics" / "current model and capabilities" / "model info"
- Any natural language request for model status or diagnostics

## Behavior

**There is no built-in function or tool named "diagnostics".** You must use the **exec** tool to run the diagnostics script. Do not tell the user that a "diagnostics" function is missing; run the command below via exec.

**Important: Be communicative. Send one short message immediately, then send an update after each section.**

0. **Reply first — before running any diagnostics**  
   Your **first** message to the user must be exactly this (copy it; do not wait to compose more):  
   **"Running diagnostics… (1) Model status (2) Config/capabilities (3) Tool and skill validation. Starting now."**  
   Send that message immediately, then run section 1. Do not add extra text to the first message; keep it short so the user sees feedback right away.

1. **Section 1 — Model status**  
   Use **exec** to run: `node /app/scripts/diagnostics/cli.js --section models_status` (or from workspace base: `node tools/diagnostics/cli.js --section models_status`).  
   Parse the JSON from stdout. **Send the user an immediate update** with the current model (defaultModel / resolvedDefault), fallbacks, and any auth overview. Do not wait for the next sections.

2. **Section 2 — Config / capabilities**  
   Use **exec** to run: `node /app/scripts/diagnostics/cli.js --section config_read`.  
   Parse the JSON from stdout. **Send the user an immediate update** with primary model, fallbacks, and capability details (context window, max tokens, input modalities) from the config. Do not wait for validation.

3. **Section 3 — Tool and skill validation**  
   Use **exec** to run: `node /app/scripts/diagnostics/cli.js --section validation`.  
   Parse the JSON from stdout. The output includes `validationTable` (or `operations[].data.validation`): an array of `{ name, type: "tool"|"skill", status }` covering **all** slash-command tools and skills (workspace tools, config skills, and OpenClaw bundled skills). Status is "Verified", "Ready", "Missing (...)", or an error message. It may also include **`skipped`**: an array of `{ name, type, reason }` for items intentionally skipped (e.g. **restart**). **Send the user an immediate update** presenting the validation as a **table**: columns **Name**, **Type**, **Status**. **If `skipped` is present and non-empty, alert the user:** e.g. "The following were skipped and were not validated: restart (skill)."

4. **Failure log info**  
   If any section run fails (tool exits non-zero or invalid JSON), report that to the user in the update for that section and include stderr or the failure log. Do not hide failures.

## Response format

- **First message only**: send exactly the short phrase above (Running diagnostics… with the three steps). No model round-trip can send a message before the model runs; this phrase minimizes latency. (True instant feedback would require a gateway system message when /diagnostics is invoked; that would be a Node-side change in the OpenClaw gateway.)
- After **each** section: send one message with that section's result.
- For the validation section: present a clear **table** with columns Name | Type | Status (Verified or error message). **If the tool output includes `skipped` (e.g. the restart skill), alert the user that those items were skipped** so they are not validated or run during diagnostics.
- Do **not** retry the tool yourself; the tool retries each operation at most 3 times and reports log info.

## Error Handling

- If a section run fails, report it in that section's update and include any stderr or `log` from the tool output.
- If the tool prints invalid JSON, say so and include the raw output you see.

## Dependencies

- Node.js 18+
- Diagnostics is a **script** run via **exec**, not a named tool. Path: `node /app/scripts/diagnostics/cli.js` (Docker) or `node tools/diagnostics/cli.js` (from workspace base). No extra env vars required.