Cloud to Ground.md

Purpose: Useful in software dev for getting a systemic analysis before implementing a feature/bug. Uses "cloud to ground" methodology (zooming in and out of the codebase to understand it more thoroughly). Applies "tracer bullet" methodology to be more direct with the implementation.

Step 1: Define the Build Target

I need to build: [FEATURE DESCRIPTION]

Step 2: Identify Tracer Bullet

Define the minimal viable technical spike:

What's the simplest end-to-end flow that delivers user value?
What's the riskiest technical assumption to validate first?
What integration points are most likely to break?

Step 3: Complexity Assessment

Rate tracer complexity (1-5):

1-2 (Simple): Basic integration, single API call → Target: 1-2 iterations
3 (Moderate): Multi-step flow, data transformation → Target: 2-3 iterations
4-5 (Complex): Real-time, multi-service coordination → Target: 3-4 iterations

Full feature complexity: [For reference only]

Step 4: Priority-Filtered Unknowns

🔍 KNOWLEDGE CONFIDENCE: [HIGH/MEDIUM/LOW] [LEARNING:]

IF LOW: "🚨 ESCALATION REQUIRED - Domain knowledge insufficient for safe analysis."

Focus only on tracer-critical unknowns:

Core integration feasibility
Critical performance constraints
Security requirements for MVP
Essential error handling
[Add specific unknowns based on feature]

Step 5: Targeted Clarification

📋 INQUIRY CONFIDENCE: [HIGH/MEDIUM/LOW] [LEARNING:]

IF LOW: "🚨 ESCALATION REQUIRED - Cannot formulate safe clarification questions."

Ask up to 5 critical questions (numbered multiple choice):

[Most critical tracer unknown] - A) Option B) Option C) Option
[Second priority] - A) Option B) Option C) Option
[Third priority] - A) Option B) Option C) Option [Continue up to 5 questions max]

Step 6: Confidence Assessment

After user responds, provide:

IMPLEMENTATION CONFIDENCE: [HIGH/MEDIUM/LOW] (X%) [LEARNING:]

HIGH (80%+): "With these answers, I can determine [aspects] with high certainty."
MEDIUM (50-79%): "⚠️ FLAGGED UNCERTAINTIES: [specific gaps]"
LOW (<50%): "🚨 MANDATORY HUMAN REVIEW - Insufficient knowledge to proceed safely."

Step 7: Tracer Iterative Refinement

Each iteration expands tracer systematically:

Iteration Structure:

TRACER FOCUS: What does this iteration prove/expand?
CLOUD: User value delivered at this tracer level
GROUND: Implementation for current tracer scope only
VALIDATION CONFIDENCE: [HIGH/MEDIUM/LOW] [LEARNING:] - IF LOW: "🚨 ESCALATION REQUIRED"
VALIDATION: Does tracer work? What's next expansion priority?
EXPANSION DECISION: Ready to expand scope or strengthen current tracer?

Iteration Progression:

Iteration 1: Prove core integration works (happy path only)
Iteration 2: Add essential error handling
Iteration 3: Address critical security/performance requirements
Iteration N: Expand toward full feature scope

Step 8: Dynamic Continuation Rules

CONTINUE IF: Critical tracer unknowns remain unresolved
AUTO-ESCALATE: If unknowns remain at target limit, reassess complexity and justify extension
HARD STOP: 6 iterations maximum
STOP WHEN: Tracer proves viability and path to full implementation is clear

Step 9: Accuracy Flywheel & Final Delivery

Once tracer is validated:

📈 ACCURACY FLYWHEEL:

Abstentions: [What "I don't know" moments occurred? What should be researched?]
Corrections: [What predictions were wrong? What patterns emerged?]
Next Session: [One thing to watch for or do differently next time]

Implementation Plan (Markdown): Complete development roadmap including:

Tracer bullet findings and validated assumptions
Step-by-step implementation sequence from tracer to full feature
Technical architecture decisions discovered through iterations
Integration points and API specifications identified
Security, performance, and error handling requirements uncovered
Risk mitigation strategies for each development phase
Testing approach based on tracer validation
Deployment considerations and rollout strategy

Show iteration count, unknowns resolved, and expansion reasoning throughout process.

iamhenry/Cloud to Ground.md