| name | ai-brainstorm |
| description | Forces rigorous design interrogation BEFORE any code: explores approaches, surfaces ambiguity, gathers evidence, produces an approved spec that becomes the contract for /ai-plan. Trigger for 'lets add X', 'how should we handle Y', 'whats the best approach', 'I am thinking about', 'what should we build for'. Not for existing approved specs; use /ai-plan instead. Not for execution; use /ai-build instead. |
| effort | mid |
| argument-hint | [feature or problem description] [work item ID] | --consolidate-spec <slug> |
| model_tier | sonnet |
| mirror_family | antigravity-skills |
| generated_by | ai-eng sync |
| canonical_source | .claude/skills/ai-brainstorm/SKILL.md |
| edit_policy | generated-do-not-edit |
Brainstorm
Design-interrogation skill: forces rigorous thinking BEFORE any code and produces an approved spec — the contract for /ai-plan.
Workflow
Principle: §10.6 SDD — this skill produces the spec. HARD GATE: no implementation runs until the operator explicitly approves it.
- -1. Live-slot guard (D-167-05) — UNLESS
--consolidate-spec, run python .ai-engineering/scripts/spec_lifecycle.py slot_status BEFORE any spec.md write (Step 6 and the Step 0b condensed path both write). If JSON reports occupied: true AND state ≠ shipped: surface spec_id/slug/state and ask the operator to consolidate first (/ai-brainstorm --consolidate-spec <slug>) or confirm the overwrite. On occupied: false, shipped, or any script error: proceed silently. Fail-open — advisory, never blocks interrogation.
- 0a. Fast-path
--consolidate-spec <slug> — read .agents/skills/_shared/consolidate-spec.md and run the shared handler (resolve record, append _history.md row via spec_lifecycle.py mark_shipped, clear spec.md/plan.md to placeholders). STOP after consolidation — no interrogation. Fail-open on missing script.
- 0b. Auto-spec gate — follow
handlers/auto-spec-gate.md. Classify the working-tree diff via ai_engineering.brainstorm.auto_spec_gate.classify_diff BEFORE interrogation. Any hard trigger OR any threshold exceeded → full interrogation (Step 1). Else → condensed-spec path: draft the minimum-viable spec from the diff, present for one-shot approval, STOP at Step 7. manifest.brainstorm.auto_spec_gate.enabled: false → no-op (fail-open to full interrogation).
- 0. Lifecycle bootstrap (before evidence sweep) —
python .ai-engineering/scripts/spec_lifecycle.py start_new <slug> <title> mints/refreshes the DRAFT record at .ai-engineering/state/specs/<slug>.json. Fail-open: non-zero exit logs and continues.
- 1. Work-item context (only when a work-item ID is given, e.g.
AB#100 / #45):
a. read manifest.yml work_items for active provider + team config;
b. fetch item + hierarchy — GitHub gh issue view <n> --json title,body,labels,milestone,assignees; Azure az boards work-item show --id <n> --expand relations -o json;
c. walk Feature → User Story → Tasks (parent/child relations);
d. use all standard + custom fields the platform provides;
e. pre-fill refs in the generated spec frontmatter;
f. /ai-board sync refinement <ref> to move the item to refinement (fail-open).
- 2. Enhance input — follow
handlers/prompt-enhance.md to clarify and sharpen user input before interrogation.
- 3. Evidence sweep — when current state spans multiple repo / governance surfaces, dispatch parallel read-only
ai-explore passes first, summarize, then sharpen the next question or spec boundary.
- 4. Interrogate — follow
handlers/interrogate.md (see Questioning Rules).
- 5. Propose approaches — 2-3 options with trade-offs (never one). Scope was settled at Step 0b; no post-interrogation file-count heuristic.
- 6. Draft spec — write
.ai-engineering/specs/spec.md, then run Step 0.5 section pre-flight (deterministic, BEFORE any LLM pass): ai-eng spec verify --sections .ai-engineering/specs/spec.md (spec-139 M7) scans the five required headers (## Summary, ## Goals, ## Non-Goals, ## Decisions, ## Risks) and emits JSON valid / missing_sections / present_sections. Exit 1 → patch missing headers, re-run. Only on valid=true proceed to LLM schema validation against .ai-engineering/reference/spec-schema.md.
- 7. Board sync (ready) — if a work-item ID was given at Step 1,
/ai-board sync ready <ref> (fail-open).
- 8. Review spec — follow
handlers/spec-review.md (max 3 iterations).
- 9. STOP — approval. On operator approval, call
python .ai-engineering/scripts/spec_lifecycle.py approve <spec_id> (<spec_id> from frontmatter spec:, fallback slug:) to transition DRAFT→APPROVED and mirror status: approved into spec.md. Fail-open (D-161-08): non-zero exit logs, does NOT block STOP. Present the approval ask with the 6-field value block (Bottom line / Why it matters / What's done / Risk / Next / Details) per .ai-engineering/reference/value-lens.md at the resolved audience level (default full) — the spec text stays a carve-out. Then present the approved spec; user runs /ai-plan.
Questioning Rules
- ONE question at a time — never batch. Prefer multiple choice (A/B/C) over open-ended.
- Frame each question and option in plain terms (impact / effort / risk) per
value-lens.md so a non-technical sponsor can answer.
- Challenge vague language — "improve", "optimize", "clean up" are not requirements.
- Push back on scope creep: "Is this in scope for v1?" Explore edge cases the user has not mentioned.
- Max 10 questions per session; if you need more, the problem is too big — split it.
- When input requires research to understand current state (audits, "is this working?", "how is X organized?"): gather data, present findings, THEN interrogate. Research precedes interrogation, it is not interrogation.
Common Mistakes
- Writing implementation details in the spec — specs describe WHAT, not HOW.
Integration
Called by: user, or /ai-plan when requirements are unclear. Calls: handlers/auto-spec-gate.md (Step 0b classifier via ai_engineering.brainstorm.auto_spec_gate.classify_diff), handlers/prompt-enhance.md, handlers/interrogate.md, handlers/spec-review.md, /ai-board sync (refinement + ready), .ai-engineering/scripts/spec_lifecycle.py start_new (fail-open bootstrap) + approve (Step 9 DRAFT→APPROVED, fail-open). Transitions to: /ai-plan (ONLY — never directly to /ai-build).
Examples
User: "AB#456 says 'improve search performance' — what does that even mean?"
/ai-brainstorm AB#456
Pulls the work item, surfaces ambiguity, drives the user to specific measurable acceptance criteria, links the spec to the work item, and produces an approved spec at .ai-engineering/specs/spec.md with decisions recorded.
$ARGUMENTS
