/plan

Start a task the moo way. Will not proceed until ≥85% confident in intent.

Phase 1: Surface Unknowns

Before anything else, identify what's unclear.

Ask yourself (silently):

• What problem does this solve today?
• Who's the user and what's their journey?
• What's the simplest version that works?
• What would make this fail catastrophically?
• What constraints exist (tech, perf, compat)?
• What's been tried before?

If confidence < 85%, ask the user 3-5 targeted questions using the Ask tool:

• Open-ended (no yes/no)
• Designed to surface blind spots
• Cover: purpose, success criteria, constraints, edge cases

Wait for answers before proceeding.

Phase 1b: Prime Skills

Once intent is ≥85% confident, invoke /hope:prime to select and load relevant skills for the task.

This ensures the right workflows and tools are in context before planning begins.

Phase 2: Classify & Invert

Once intent is clear:

1. Classify: Build (A) / Debug (B) / Refactor (C)
2. Invert: List 3-5 failure modes with impact levels

## Failure Analysis
- [Mode]: [CATASTROPHIC/HIGH/MEDIUM/LOW]

Phase 3: Library Search (Build tasks)

Search production libraries:

• npm / PyPI / crates.io
• Find ≥2 options OR justify why none exist
• Evaluate: downloads, maintenance, security

Phase 4: Assess Confidence

State confidence as range:

• Apply novelty decay (first time = max 70%)
• Cite specific evidence

Confidence gates:

• < 70% → More research needed, don't recommend yet
• 70-85% → Can proceed with monitoring
• ≥ 85% → Ship with confidence

Phase 5: Present Plan

## Intent
[What we're building and why - one sentence]

## Approach: Layer 0
[Simplest working version using production library]

## Failure Modes
- [Mode 1]: [Impact] - [Mitigation]

## Success Criteria
- [How we know it works]

---
[Quality Footer - see ../skills/soul/references/quality-footer.md]

Phase 5b: Project-Level Planning (≥8 story points only)

For high-complexity tasks, add project-level controls:

### Kill Criteria
Conditions that stop the project:
1. [Specific measurable condition]
2. [Specific measurable condition]
3. [Specific measurable condition]

### Pivot Triggers
Signals requiring direction change:
1. [Observable signal]
2. [Observable signal]
3. [Observable signal]

### 48-Hour De-risking Test
Riskiest assumption: [Identify #1 risk]
Test: [Specific experiment to run]
Success criteria: [Measurable outcome]
Failure criteria: [Measurable outcome]
Pivot if fails: [Alternative direction]

Phase 6: Wait

Do not implement until user says "proceed".

If user provides feedback, incorporate and re-present plan.

---

Task

$ARGUMENTS