Feasibility Assessment Skill

Read .hackathon/STATE.md, .hackathon/IDEA.md, and .hackathon/mockups/ (latest version) first.

Soft gate: Ideally runs after mockups exist so the scope is concrete. Can run without them.

Assessment Framework

Input Parameters

From STATE.md, extract:

• Total available time (from start/end times, minus kickoff/demo/buffer)
• Number of developers
• Skill levels (from team descriptions in STATE.md)

Effort estimation calibration for agentic coding:

Work Type	Multiplier vs. Traditional	Notes
Greenfield CRUD (routes, forms, basic UI)	0.3x-0.5x (2-3x faster)	Agent handles boilerplate well
AI integration (API calls, prompt engineering)	0.8x-1.0x	Requires human judgment on prompts
Infrastructure (deploy, CI/CD, databases)	0.8x-1.0x	Config-heavy, agent helps but needs verification
Novel UI interactions (drag-drop, real-time)	0.7x-1.0x	Agents know common patterns, struggle with custom ones
Debugging & edge cases	1.0x-1.5x	Can be slower when agent-generated code has subtle bugs
Integration across subsystems	1.2x-1.5x	Integration tax: any task touching 2+ subsystems gets this penalty

The integration tax is the most commonly underestimated factor. Individual pieces working independently does NOT mean they work together. Budget explicit integration time for every connection point in the architecture.

Step 1: Decompose into Work Chunks

Break the MVP scope from IDEA.md into concrete work chunks. Each chunk gets:

Chunk	Description	Estimated Hours (with agentic coding)	Risk	Dependencies
[name]	[what it is]	[hours]	Low/Medium/High	[what it needs first]

Risk levels:

• Low: Well-documented tech, straightforward, agent can handle most of it
• Medium: Some unknowns, may need debugging, agent might need guidance
• High: Novel integration, poorly documented, likely manual debugging

Step 2: Timeline Mapping

Map the chunks onto the actual timeline:

• Account for the time already spent on planning (it's an investment, not a loss)
• Account for breaks, meals, context-switching
• Account for integration time (individual pieces working != everything working together)
• Reserve 2 hours minimum for demo prep, deployment, and buffer
• Reserve 1 hour for final testing and bug fixing

Step 3: Verdict

Produce one of:

• GO: MVP is achievable with confidence. Nice-to-haves are stretch goals.
• GO WITH CUTS: MVP is tight. Recommend specific cuts. List what to drop and what it saves.
• RESCOPE: MVP as defined is too ambitious. Propose an alternative MVP that fits.

Step 4: Scope Recommendations

For each nice-to-have from IDEA.md, assess: "If we have 2 extra hours, is this the best use of that time?" Rank them.

For the MVP, identify: "What's the single riskiest piece? What's the fallback if it doesn't work?"

Output

Write to .hackathon/FEASIBILITY.md:

---
owner: [assessor]
status: draft
created-at: [timestamp]
last-modified: [timestamp]
phase: feasibility
---

# Feasibility Assessment

## Summary
Verdict: [GO | GO WITH CUTS | RESCOPE]
Available coding time: [X] hours across [N] developers = [total] dev-hours
Estimated effort: [Y] dev-hours (with agentic coding)
Buffer: [Z] hours

## Work Breakdown
[Table from Step 1]

## Timeline
[Mapping from Step 2]

## Recommendations
[Specific cuts or adjustments]

## Risk Register
| Risk | Likelihood | Impact | Mitigation |
|------|-----------|--------|------------|

## Decision Needed
[What the team needs to decide before proceeding]

Update STATE.md, log to changelog.