Spike Investigation Workflow

Investigate: $ARGUMENTS

Agent Delegation

For deep investigations (>1 hour), delegate to the yokay-spike-runner agent for isolated execution. This keeps investigation context separate and allows the agent to write spike reports to .claude/spikes/.

Use the yokay-spike-runner agent to investigate: $ARGUMENTS
Time box: [specify hours]
Return the decision (GO/NO-GO/PIVOT/MORE-INFO) and summary to this conversation.

For quick feasibility checks (<1 hour), run inline without delegation.

Spike Discipline

Spikes answer questions, they don't build features.

Good spike questions:

• "Can we integrate with X API?"
• "Is Y library suitable for our use case?"
• "What's the best approach for Z?"

Bad spike questions:

• "Build authentication" (that's a feature)
• "Investigate the codebase" (too vague)
• "Make it faster" (needs specific question)

Steps

1. Clarify the Question

Reframe $ARGUMENTS as a specific, answerable question.

If vague, ask:

• What decision does this inform?
• What would a "yes" answer look like?
• What would a "no" answer look like?

2. Set Time Box

Duration	Use Case
1h	Quick feasibility check
2h	Standard investigation (default)
4h	Complex evaluation
8h	Deep dive (requires justification)

3. Create Spike Task

npx @stevestomp/ohno-cli create "Spike: $QUESTION" -t spike
npx @stevestomp/ohno-cli start <task-id>

4. Define Success Criteria

Before starting, document:

## Spike: [Question]
**Time Box**: [hours]
**Started**: [timestamp]
**Must Conclude By**: [timestamp + time-box]

### Success Criteria
- [ ] [Specific criterion 1]
- [ ] [Specific criterion 2]

5. Investigate

Focus on answering the question:

• Research (docs, examples, community)
• Proof-of-concept code (if needed)
• Document findings as you go

6. Checkpoint at 50%

At half time, assess:

## Spike Checkpoint (50%)
**Progress**: [summary of findings so far]
**On Track?**: Yes / No / Pivoting
**Remaining Time**: [hours]

If off track:

• Should we narrow scope?
• Is more time needed? (requires justification)

7. Conclude with Decision

Even if time runs out, produce decision:

Decision	Meaning	Next Step
GO	Feasible, proceed	Create implementation tasks
NO-GO	Not feasible	Document why, close spike
PIVOT	New question	Create new spike task
MORE-INFO	Insufficient data	One follow-up spike max

8. Generate Spike Report

Create .claude/spikes/[date]-[slug].md:

# Spike: [Question]

**Date**: [date]
**Time spent**: [hours]
**Decision**: [GO/NO-GO/PIVOT/MORE-INFO]

## Question
[The specific question investigated]

## Answer
[Clear answer with rationale]

## Findings
[Key discoveries]

## Evidence
[Code samples, links, screenshots]

## Recommendation
[What to do next]

## Follow-up
- [ ] [Action item 1]
- [ ] [Action item 2]

9. Create Follow-up Tasks

If decision is GO:

npx @stevestomp/ohno-cli create "[implementation task]" -t feature

10. Complete Spike

npx @stevestomp/ohno-cli done <task-id> --notes "Decision: X. See .claude/spikes/[file]"

Output

## Spike Complete

**Question**: [question]
**Answer**: [GO/NO-GO/PIVOT/MORE-INFO]
**Time**: [actual] / [budgeted]
**Report**: .claude/spikes/[file].md

### Summary
[2-3 sentence summary]

### Next Steps
[Follow-up tasks created, or "No action needed"]

Related Commands

• /pokayokay:research - Longer multi-day research
• /pokayokay:work - Implement spike findings
• /pokayokay:quick - Non-investigative quick work

Examples

/pokayokay:spike Can we use Clerk for authentication?
/pokayokay:spike What's the best state management for our React app?
/pokayokay:spike Is our current architecture suitable for real-time features?