Back to Skills

exploration-debug

verified

Interactive debugging through conversation. HOUSTON guides through 4 phases, asks questions, and suggests agents for investigation.

View on GitHub

Marketplace

space-agents

thebrownproject/space-agents

Plugin

space-agents

Repository

thebrownproject/space-agents

skills/exploration-debug/SKILL.md

Last Verified

January 25, 2026

Install Skill

Select agents to install to:

Scope:
npx add-skill https://github.com/thebrownproject/space-agents/blob/main/skills/exploration-debug/SKILL.md -a claude-code --skill exploration-debug

Installation paths:

Claude
.claude/skills/exploration-debug/
Powered by add-skill CLI

Instructions

# /exploration-debug - Interactive Debugging

Debug issues through conversation. This is collaborative investigation, not a checklist. You guide through the 4 phases, ask questions, have opinions, and work toward root cause together.

## The Process

1. **Acknowledge issue, ask first question** - Don't jump to solutions
2. **Multi-round dialogue through 4 phases** - Adapt pace to complexity
3. **Suggest agents when useful** - "Want me to send an agent to trace X while we talk?"
4. **If agent spawned** - Continue talking, check results between questions with `TaskOutput block: false`
5. **Weave in results naturally** - Brief summaries, not dumps
6. **When root cause found** - Offer choice: fix now OR create bug Bead

## The Four Phases

Guide conversation through these phases. Don't rush - some issues need multiple rounds per phase.

### Phase 1: Understand the Issue

Ask about:
- What exactly is happening? (symptoms, error messages)
- Can you reproduce it? How consistently?
- What changed recently? (code, config, environment)
- When did it start?

**Red flags to watch for:**
- User already has a fix in mind - slow down, verify root cause first
- "It's probably X" - investigate, don't assume

### Phase 2: Gather Evidence

Ask about or investigate:
- Exact error messages and stack traces
- Logs around the failure point
- State of system when it fails
- What works vs what doesn't

**Suggest agent when:** You need to trace code paths, check logs, or gather evidence while continuing conversation.

### Phase 3: Find Root Cause

Ask about:
- Where does bad data originate?
- What's different between working and broken cases?
- Are there similar patterns elsewhere that work?

**Key principle:** Trace backward from symptom to source. Fix at source, not symptom.

### Phase 4: Resolution

When root cause is confirmed, ask user:

```
AskUserQuestion:
  "We've found the root cause. How do you want to proceed?"
  Options:
  - "Fix it now" - Apply the fix in this session
  - "Creat

Validation Details

Front Matter
Required Fields
Valid Name Format
Valid Description
Has Sections
Allowed Tools
Instruction Length:
3591 chars