feat: add Phase 0 (Align on Verification Criteria) to systematic-debugging

[jacksang6]

Summary

Adds a new Phase 0: Align on Verification Criteria to the systematic-debugging skill, before the existing Phase 1 (Root Cause Investigation).

Motivation

The current 4-phase process starts at "Root Cause Investigation" but skips a critical prerequisite: aligning with your human partner on what you are actually verifying. Without this alignment:

• You write tests that verify the wrong thing
• You investigate without knowing when to stop
• You waste rounds on misaligned expectations

This pattern emerged from repeated debugging sessions where the AI jumped to writing tests before confirming what "pass" meant with the human partner, leading to false RED→GREEN cycles and wasted investigation.

Changes

Section	Change
Phase 0	New phase: Define target dimension → Establish expected baseline → Define "success" → Get sign-off
Title	"Four Phases" → "Five Phases"
Red Flags	+ "Writing test code before aligning on what pass means"
Partner Signals	+ "Grill first / Let's align first" → Phase 0
Rationalizations	+ "I know what to verify, no need to align"
Quick Reference	+ Phase 0 row
Real-World Impact	+ "With Phase 0 alignment: Near-zero wasted investigation rounds"

Testing

Applied in real debugging sessions. Phase 0 alignment reduced investigation from multiple wasted rounds to a single targeted round.

Design Decisions

• Phase numbering: Uses "Phase 0" (not "Phase 1.5") because this step is a prerequisite gate — it must complete before any investigation begins, similar to how Phase 0 in other frameworks means "pre-requisite alignment"
• Sign-off requirement: Explicit "do NOT proceed to Phase 1" guard, matching the existing Iron Law pattern
• Concrete examples: Uses generic e-commerce/database examples rather than domain-specific ones, to keep the skill broadly applicable