ghm-id-register

ID Register

Validate and register new Source of Truth IDs with cross-reference integrity checks.

Workflow Overview

1. Validate Format → Check ID follows [PREFIX]-[3-digit] pattern
2. Check Uniqueness → Ensure ID doesn't already exist
3. Verify Cross-Refs → All referenced IDs must exist
4. Register Entry → Add to appropriate SoT file

Core Output Template

Element	Definition	Evidence
ID	Unique identifier	BR-101, UJ-045, API-012
Title	Short descriptive name	Clear, specific
Cross-References	Links to related IDs	All referenced IDs exist
Status	Current state	Draft / Active / Deprecated

ID Format Reference

Prefix	Domain	File
BR-	Business Rules	SoT/SoT.BUSINESS_RULES.md
UJ-	User Journeys	SoT/SoT.USER_JOURNEYS.md
API-	API Contracts	SoT/SoT.API_CONTRACTS.md
CFD-	Customer Feedback	SoT/SoT.customer_feedback.md

Step 1: Validate Format

Check ID follows the pattern:

[PREFIX]-[XXX]

Where:

• PREFIX = BR, UJ, API, or CFD
• XXX = 3-digit number (zero-padded)

Checklist

• Prefix is valid (BR, UJ, API, CFD)
• Number is 3 digits
• Format matches [A-Z]+-[0-9]{3}

Step 2: Check Uniqueness

1. Read target SoT file
2. Extract all existing IDs of same prefix
3. Verify new ID doesn't exist
4. If auto-assigning: use highest existing + 1

Checklist

• Target SoT file read
• Existing IDs enumerated
• New ID is unique

Step 3: Verify Cross-References

For each ID referenced in the new entry:

1. Identify the prefix
2. Check that ID exists in its SoT file
3. Flag any missing references

Checklist

• All BR-XXX references exist in BUSINESS_RULES
• All UJ-XXX references exist in USER_JOURNEYS
• All API-XXX references exist in API_CONTRACTS
• All CFD-XXX references exist in CUSTOMER_FEEDBACK
• Each cross-reference includes a relationship type (see references/cross-reference-patterns.md)
• Relationship types match the directional hierarchy (vertical types for cross-layer, lateral types for same-layer)

Step 3.5: Evaluate Confidence (NEW)

Before registering, assign a confidence score (1-5) based on evidence strength:

Score	Evidence Level	Examples
1/5	Assumption / PM decision	"We think users want X"
2/5	Secondary research	Competitive analysis, market report
3/5	Direct feedback	User interviews (3-5 conversations)
4/5	Validated behavior	Beta testing, small-scale usage
5/5	Production evidence	Real usage data at scale

Question to ask: What's the highest evidence supporting this entry right now? What would move it to the next confidence level?

Example confidence annotations:

• confidence: 2/5, source: competitive-analysis
• confidence: 3/5, source: 5-user-interviews-jan-2026
• confidence: 4/5, source: beta-cohort-validation

See .claude/skills/PRINCIPLES.md for detailed confidence model by SoT type.

Checklist

• Confidence score assigned (1-5)
• Highest evidence source identified
• Forward path identified ("would move to X/5 if...")

Step 4: Register Entry

Add formatted entry to SoT file:

### [ID]: [Title]

**Status**: Draft
**Created**: YYYY-MM-DD
**Confidence**: [1-5]/5 (source: [evidence source])
**Next Confidence Target**: [What would move this to next level]
**Cross-References**: [List of related IDs]

[Description]

**Acceptance Criteria**:
- [ ] Criterion 1
- [ ] Criterion 2

Example entry with confidence:

### CFD-042: Users want dark mode

**Status**: Active
**Created**: 2026-02-01
**Confidence**: 3/5 (source: 5-user-interviews-jan-2026)
**Next Confidence Target**: 4/5 (would require beta cohort validation)
**Cross-References**: FEA-008 (dark mode feature)

During interviews, 4 of 5 users mentioned desire for dark mode. Competitors (Notion, Linear, Figma) all have it.

**Acceptance Criteria**:
- [ ] Feature FEA-008 delivered to beta cohort
- [ ] Track usage: % of beta users enabling dark mode

Quality Gates

Pass Checklist

• ID format is valid
• ID is unique within its domain
• All cross-references resolve
• Entry follows SoT template
• Confidence score assigned (1-5) with source documented
• Next confidence target identified

Testability Check

• ID can be searched and found
• Cross-references are bidirectional (if required)
• Confidence score is honest (reflects actual evidence, not wishful thinking)

Anti-Patterns

Pattern	Example	Fix
Duplicate ID	Creating BR-101 when it exists	→ Check uniqueness first
Orphan reference	References UJ-999 that doesn't exist	→ Verify all cross-refs
Wrong prefix	Using BR- for an API contract	→ Match prefix to domain
Missing zero-pad	BR-5 instead of BR-005	→ Always use 3 digits
Inflated confidence	Assigning 4/5 to a PM assumption	→ Be honest about evidence level
No confidence source	"confidence: 3/5" with no source	→ Always record source (CFD-001, user-interview-jan, etc.)
Missing confidence target	Confidence assigned but no forward path	→ Ask "what would move this to 4/5?"

Boundaries

DO:

• Format validation
• Uniqueness checks
• Cross-reference verification
• Entry formatting

DON'T:

• Content decisions about ID meaning
• Approve/reject based on business logic
• Modify existing IDs

Handoff

After ID registration:

• New ID is in SoT file
• Cross-references are valid
• EPIC Execution Plan updated with new ID
• Ready for implementation