write-architecture-spec

Overview

Runs IA specification authoring in two shards: design → deepen.

Prerequisites

1. Decomposition outputs exist (.memory/wiki/specs/ia/ skeletons + indexes)
2. .memory/pipeline/progress/spec-pipeline.md exists and includes IA column states
3. Stack map is resolved for required categories (Databases, Security, Surfaces)

Step-by-Step

Step 0 — Pipeline-state shard selection

1. Read .memory/pipeline/progress/spec-pipeline.md.
2. Select lowest-numbered IA shard with not-started status unless override is explicitly requested.
3. If no IA shard remains: stop and direct next step to /write-be-spec.

Step 1 — Run design shard

Call skill: write-architecture-spec-design

Expected outputs:

• Fully drafted IA sections for selected shard:
  • interactions
  • contracts
  • data models
  • access control
  • accessibility
  • event schemas
  • edge cases

Step 2 — Run deepen shard

Call skill: write-architecture-spec-deepen

Expected outputs:

• Iterative deepening passes complete
• IA shard finalized and verified
• IA index updated for shard completion
• spec pipeline IA state updated
• ambiguity gates executed

Quality Gate

Do not close workflow unless all pass:

• complete interaction flows
• typed contracts with explicit errors
• complete data model relationships/constraints
• role coverage across all interactions
• edge-case coverage for concurrency/deletion/conflicts
• bidirectional cross-shard dependencies

Completion Checklist

• target shard selected from pipeline state
• design shard completed and user-reviewed
• deepen shard completed and user-reviewed
• IA index updated
• pipeline tracker updated
• ambiguity gates completed
• next valid step recommended based on remaining IA shards

Next Steps

• If IA shards remain: /write-architecture-spec
• If IA is complete and /audit-ambiguity ia is clean: /write-be-spec