ai-brainstorm

# Brainstorm ## Purpose Design interrogation skill. Forces rigorous thinking BEFORE any code is written. Produces an approved spec that becomes the contract for `/ai-plan`. HARD GATE: this skill produces a spec. No implementation happens until the user explicitly approves it. ## When to Use - User says "I want to build...", "how should we...", "let's design..." - New feature, architecture change, or ambiguous requirement - The spec depends on evidence scattered across multiple repo or governance surfaces - Any work where jumping straight to code would be premature ## Process -1. **Live-slot guard** (D-167-05) — UNLESS invoked with `--consolidate-spec`, run `python .ai-engineering/scripts/spec_lifecycle.py slot_status` BEFORE any `spec.md` write (Step 6 and the Step 0b condensed-spec path both write it). If the JSON reports `occupied: true` AND `state` is not `shipped`, the slot already holds an un-shipped spec — surface its `spec_id`, `slug`, and `state` and ask the operator to either consolidate it first (`/ai-brainstorm --consolidate-spec <slug>`) or explicitly confirm the overwrite. On `occupied: false`, a `shipped` state, or any script error, proceed silently. **Fail-open**: this guard is advisory and never blocks interrogation — a non-zero exit logs and continues. 0a. **Fast-path — `--consolidate-spec <slug>`**: when invoked with the `--consolidate-spec` flag, read `.agents/skills/_shared/consolidate-spec.md` and execute the shared...