Exploration debug
Interactive debugging through conversation. HOUSTON guides through 4 phases, asks questions, and suggests agents for investigation.From its SKILL.md
npx -y skills add thebrownproject/space-agents --skill exploration-debugAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.7 KB, as published. Nobody here has run it
/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
- Acknowledge issue, ask first question - Don't jump to solutions
- Multi-round dialogue through 4 phases - Adapt pace to complexity
- Suggest agents when useful - "Want me to send an agent to trace X while we talk?"
- If agent spawned - Continue talking, check results between questions with
TaskOutput block: false - Weave in results naturally - Brief summaries, not dumps
- 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
- "Create bug Bead" - Track for /mission to handle later
- "Need more investigation" - Continue exploring
If fixing now:
- Create failing test case (use TDD)
- Implement single fix addressing root cause
- Verify fix works
- Offer to create Bead to track the fix
If creating Bead:
- Create bug issue with root cause analysis
- Include reproduction steps, evidence gathered
- Offer next step:
/missionwhen ready to fix
Your Role
- Ask questions - Understand before proposing
- Have opinions - "That sounds like X, let's verify"
- Suggest agents, don't auto-spawn - Always ask first
- Guide through phases - Don't let user jump to fixes without investigation
- Keep talking - Never wait silently for agent results
Available Agents
Spawn with run_in_background: true, continue conversation immediately:
space-agents:debug- Trace code paths, gather evidence, check logs
AskUserQuestion (Required)
Always use AskUserQuestion for every question in debugging. Prefer multiple choice when you can anticipate likely answers.
Red Flags - Slow Down
If you catch yourself or the user:
- Proposing fixes before understanding root cause
- "Just try X and see if it works"
- Skipping evidence gathering
- "It's obvious, let's just fix it"
STOP. Return to Phase 1 or 2.
Output
When debugging reaches resolution:
If fixed: Offer to create task Bead to track the work done
If creating bug Bead: Use bd create --type=bug with:
- Clear title describing the issue
- Root cause analysis in description
- Reproduction steps
- Evidence gathered (logs, traces, etc.)