Troubleshooting Guide Builder

What this skill does

This skill builds a structured troubleshooting guide that maps symptoms to causes to fixes for any software tool, API, or system. Each entry follows a consistent format: what the user experiences, why it happens, and exactly how to fix it — including specific commands, configuration changes, or code snippets. The result is a guide that developers or support staff can scan quickly to resolve issues without digging through source code or guessing.

Use this when writing documentation for a new tool, after an incident reveals a common failure mode, when your support team keeps answering the same questions, or when preparing runbooks for operational issues.

How to use

Claude Code / Cline

Copy this file to .agents/skills/troubleshooting-guide-builder/SKILL.md in your project root.

Then ask:

• "Use the Troubleshooting Guide Builder skill to write a troubleshooting guide for our CLI tool."
• "Build a troubleshooting guide for the most common deployment errors using the Troubleshooting Guide Builder skill."

Provide:

• A description of the tool, API, or system
• A list of known issues, error messages, or symptoms (ideally from real support cases)
• Any relevant error codes, configuration options, or environment details

Cursor

Add the instructions below to your .cursorrules or paste them into the Cursor AI pane. Provide the tool description and known issues.

Codex

Paste the tool description, error messages, and any existing support notes. Ask Codex to follow the instructions below.

The Prompt / Instructions for the Agent

When asked to build a troubleshooting guide, follow these steps:

Step 1 — Gather known issues

If given a list of symptoms or errors, use them as the basis. If not, generate a comprehensive list of likely issues by thinking through:

• Installation and setup failures
• Authentication and permissions errors
• Configuration mistakes
• Rate limits and quota errors
• Network and connectivity issues
• Common misunderstandings about how the tool works
• Environment-specific issues (OS, version, dependencies)
• Data format or validation errors

Step 2 — For each issue, define the three parts

Every troubleshooting entry must have:

1. Symptom — What does the user actually see? Use the exact error message if possible. Be specific: "The command exits with code 1 and prints Error: ECONNREFUSED" is better than "it doesn't work."

2. Cause — Why does this happen? This is the technical explanation. Keep it brief (1–3 sentences) but accurate enough that an informed reader understands the underlying mechanism.

3. Fix — The exact steps to resolve it. Use numbered steps for multi-step fixes. Include specific commands, file paths, config values, or code snippets. The fix must be actionable — not "check your configuration" but "open ~/.config/tool.json and ensure timeout is set to a value greater than 0."

Step 3 — Add diagnostic steps where needed

For issues where the fix isn't the same for everyone, add a diagnostic section before the fix:

**How to diagnose**: Run `tool --verbose` and look for the line starting with "Connection". If it shows "refused", the server is not running. If it shows "timeout", the server is running but unreachable.

Step 4 — Organize by category

Group issues into logical categories. Common categories:

• Installation & Setup
• Authentication & Permissions
• Configuration
• Network & Connectivity
• Rate Limits & Quotas
• Data & Input Errors
• Performance
• Common Misunderstandings

Step 5 — Format the guide

# Troubleshooting Guide — [Tool/System Name]

## Quick Reference

| Symptom | Section |
|---------|---------|
| [Short symptom description] | [Category] |
| ... | ... |

---

## [Category Name]

### [Short symptom title]

**Symptom**
[Exact error message or description of what the user sees]

**Cause**
[Brief technical explanation of why this happens]

**Fix**
1. [Step 1]
2. [Step 2]

[or for single-step: just the fix as a paragraph or code block]

**Still not working?**
[Optional: what to check next or where to get help]

---

[Repeat for each issue]

Step 6 — Include a "Before You Start" section

At the top, add a short section with universal first steps that resolve many issues:

• Check the current version: tool --version
• Check the logs: where are they?
• Enable verbose mode: how?
• Verify the environment: key environment variables to check

Example

Input to Agent:

"Use the Troubleshooting Guide Builder skill for our CLI deployment tool called deploytool. Known issues from support:

1. Deploy hangs forever
2. 'Invalid credentials' error even with correct API key
3. 'Region not found' error
4. Deployed version doesn't match local version"

Output from Agent:

Troubleshooting Guide — deploytool

Before You Start

Run these commands before diving into specific issues:

deploytool --version          # Check you're on the latest version
deploytool config show        # Review current configuration
deploytool deploy --verbose   # Enable verbose logs for your next deploy

---

Authentication

"Invalid credentials" error even with the correct API key

Symptom

Error: Invalid credentials. Check your DEPLOYTOOL_API_KEY.

Cause The API key is read from the DEPLOYTOOL_API_KEY environment variable. If the variable is set in a .env file but the file isn't loaded, or if the variable name has a typo, the tool falls back to an empty string and authentication fails.

Fix

1. Confirm the variable is set: echo $DEPLOYTOOL_API_KEY
2. If empty, export it: export DEPLOYTOOL_API_KEY=your_key_here
3. Verify it's the correct key in your dashboard
4. If using a .env file, ensure your shell loads it: source .env

---

Deployment Issues

Deploy hangs and never completes

Symptom The deploytool deploy command starts but never finishes, even after several minutes.

Cause The tool waits for a health check response from the deployed service. If the service fails to start or the health check endpoint isn't reachable, the deploy hangs until a 10-minute timeout.

How to diagnose Run with --verbose and look for "Waiting for health check at...". Navigate to that URL in a browser — if it returns an error, your application failed to start.

Fix

1. Check your application logs: deploytool logs --tail 50
2. Fix the application error causing startup failure
3. Re-deploy: deploytool deploy

If your service intentionally has no health check: deploytool deploy --no-health-check

Notes

• Use exact error messages in the Symptom section — users search for the literal error text they see.
• The Fix section should be written so a tired developer can follow it at 2am without thinking. Be explicit.
• Update this guide after every incident where an issue wasn't already covered.