reconcile¶
Generate a unified execution spec from approved specifications.
Usage¶
Description¶
The reconcile command combines all approved specifications into a unified execution spec (spec.md). This is the final step before exporting to AI coding agent execution systems.
The reconciliation process:
- Verifies all required specs are approved
- Detects conflicts between specs
- Resolves conflicts using LLM-assisted analysis
- Generates unified
spec.md - Creates
spec.eval.jsonwith reconciliation metadata
Prerequisites¶
All required specs must be approved before reconciliation:
Conflict Detection¶
The reconciler detects conflicts across specs:
| Conflict Type | Specs Involved | Example |
|---|---|---|
| Performance vs Feature | PRD, TRD | Feature requires real-time sync but TRD specifies batch processing |
| Security vs Usability | UXD, TRD | UX flow bypasses security controls |
| Scope vs Timeline | MRD, PRD | PRD includes features not in MRD scope |
| Missing Traceability | PRD, TRD | Requirements not covered in technical design |
Output Files¶
spec.md - Unified execution specification containing:
- Executive summary
- Consolidated requirements
- Technical architecture
- Implementation phases
- Acceptance criteria
eval/spec.eval.json - Reconciliation metadata:
{
"spec_type": "spec",
"generated_at": "2024-01-15T10:30:00Z",
"sources": ["mrd", "prd", "uxd", "trd", "ird"],
"conflicts": [
{
"id": "C001",
"type": "performance_vs_feature",
"description": "Real-time requirement conflicts with batch architecture",
"severity": "high",
"resolution": "Implement hybrid approach with real-time for critical paths"
}
],
"decision_log": [...],
"status": "reconciled_with_tradeoffs"
}
Reconciliation Status¶
| Status | Description |
|---|---|
reconciled |
No conflicts detected |
reconciled_with_tradeoffs |
Conflicts resolved with documented tradeoffs |
needs_review |
Unresolved high-severity conflicts |
Examples¶
# Run reconciliation
visionspec reconcile
# Output:
# ✓ All required specs approved: [mrd prd uxd trd ird]
# ⋯ Reconciling specs...
# ✓ Generated docs/specs/my-project/spec.md
# Sources: [mrd prd uxd trd ird]
# Conflicts detected: 2
# ✓ C001: Performance vs feature trade-off
# ✓ C002: Security model alignment
# ✓ Generated docs/specs/my-project/eval/spec.eval.json
Handling Failures¶
If reconciliation fails due to missing approvals:
If high-severity conflicts cannot be resolved, the status will be needs_review and manual intervention is required.