---

name: troubleshooting description: A robust troubleshooting framework. Use this skill anytime the user reports something isn't working, is buggy, or is throwing errors.

Troubleshooting

This skill helps identify when you're applying a bandaid fix versus addressing root causes, and when to escalate to deeper investigation.

Core Principles

1. Bandaid Detection

Before implementing any fix, check if it's treating symptoms vs. root cause:

Red flags indicating bandaid fixes:

• Adding try/catch blocks that hide errors without understanding why they occur
• Implementing timeouts or retries without investigating why failures happen
• Duplicating logic to work around a broken component
• Increasing resource limits without understanding why resources are exhausted
• Caching to hide performance issues without addressing underlying inefficiency
• Adding null checks without understanding why nulls appear
• Hard-coding values that should be dynamic (IDs, paths, credentials, configuration values)

When you detect a bandaid: "I could [quick fix], but that just masks the real issue: [root cause]. To fix properly, I need to [proper solution / missing information]. Should I implement the workaround for now or investigate the root cause?"

2. Systemic Issue Detection

Watch for signs that a bug indicates larger problems:

Indicators of systemic issues:

• Same type of error occurring in multiple places
• Issue requires workarounds in multiple locations
• Root cause points to architectural decisions
• Fix would require changing fundamental assumptions
• Similar issues have been "fixed" before with workarounds

When you detect systemic issues: "This appears to be a symptom of a larger issue: [systemic problem]. The immediate fix is [X], but this suggests we should also consider [architectural change]."

3. Escalation to Deep Investigation

For most issues, attempt a straightforward fix. If any of these occur, you MUST read references/collaborative-workflow.md and follow that process:

Escalation triggers:

• Simple fix attempt fails or reveals complexity
• Multiple possible root causes exist
• Issue is more complex than it initially appeared
• You're uncertain whether a solution is proper or a bandaid
• Investigation requires information you don't have access to

When triggered, immediately use the view tool:

Copyview references/collaborative-workflow.md

Then follow the detailed investigation process described in that file.

Quick Troubleshooting

For straightforward issues, proceed autonomously:

1. Identify the issue - Read error messages, examine code, check logs
2. Verify it's not a bandaid - Check against red flags above
3. Implement the fix - Address the root cause
4. Verify - Confirm the specific symptom is resolved

If at any point this becomes unclear or the fix would be a bandaid, escalate to the collaborative workflow.

Anti-Patterns

Assumption-driven debugging:

• Don't assume you understand the architecture
• Don't guess at configuration or environment details
• Don't implement solutions based on incomplete information

Premature solutions:

• Don't propose fixes before understanding the root cause
• Don't implement the first solution that comes to mind without evaluation
• Don't skip verification that the fix actually addresses the root cause