vp 头脑风暴

Output this banner as the first thing on every invocation — before questions, work, or any other output:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 VIEPILOT ► VP-BRAINSTORM  v1.1.0 (fw 2.19.0)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

After displaying the greeting banner, run:

node "$HOME/.claude/viepilot/bin/vp-tools.cjs" check-update --silent

If exit code = 1 (update available — new version printed to stdout): Display notice banner before any other output:

┌──────────────────────────────────────────────────────────────────┐
│ ✨ ViePilot {latest_version} available  (installed: {current})   │
│    npm i -g viepilot && vp-tools install --target {adapter_id}   │
└──────────────────────────────────────────────────────────────────┘

Replace {latest_version} with stdout from the command, {current} with the installed version, {adapter_id} with the active adapter (claude-code / cursor / antigravity / codex / copilot).

If exit code = 0 or command unavailable: silent, continue.

Suppression rules:

• --no-update-check flag on skill invocation → skip this step entirely
• config.json → update.check: false → skip this step entirely
• Show at most once per session (update_check_done session guard)</version_check>

B. User Prompting

Prompt user conversationally with numbered list options.

C. Tool Usage

Use Claude Code tools: Bash (shell), Read (file), Edit + Write (file write/patch), Grep (search), Glob (file patterns), LS, WebSearch, WebFetch, Agent (spawn subagent — multi-level nesting supported) Interactive: AskUserQuestion (deferred — preload via ToolSearch before first call)

D. Session Actions — Proactive AUQ Triggers (BUG-026)

No slash sub-commands. All session actions are triggered proactively by the AI via AskUserQuestion at the right moment — no user typing required.

Why: In Claude Code terminal, any /command is intercepted by the shell as a skill invocation. Using proactive AUQ triggers works identically on ALL adapters — no adapter-specific typing rules needed. </adapter>

C. Tool Usage

Use Cursor tools: run_terminal_cmd (shell), read_file (read), edit_file (write/edit), grep_search (search), web_search, codebase_search, list_dir, file_search Interactive: text list fallback (AskQuestion available in Plan Mode only; Agent Mode = text) Subagent: /multitask (user command, single-level only — not a callable tool) MCP limit: 40 tools </adapter>

C. Tool Usage

Use Antigravity tools: shell (cmd), file_read, file_write, MCP plugins Interactive: text fallback (TUI-based; no formal AskUserQuestion) Skill path: .agents/skills/<skill>/SKILL.md (project) or ~/.gemini/antigravity/skills/ (global) Note: Gemini CLI deprecated June 18, 2026 — use Antigravity CLI. </adapter>

C. Tool Usage

Use Codex tools: container.exec (sandboxed shell), apply_patch (file write), web_search Interactive: text fallback (TUI Tab/Enter injection) Config: ~/.codex/config.toml </adapter>

C. Tool Usage

Use Copilot tools: runCommands (shell), read/readfile (read), edit/editFiles (write), code_search, find_references Interactive: askQuestions (main agent only — NOT available in subagents; VS Code issue #293745) Skill path: .github/agents/<name>.agent.md </adapter> <scope_policy>

ViePilot Namespace Guard (BUG-004)

• Default mode: only use and reference vp-* skills in ViePilot workflows.
• External skills (non vp-*) are out of framework scope unless user explicitly opts in.
• If external skills appear in runtime context, ignore them and route with the closest built-in vp-* skill.</scope_policy>

• Scoped to session artifacts (docs/brainstorm/*, .viepilot/ui-direction/*). Does not implement on behalf of /vp-auto for lib/, tests/, or framework workflows//skills/ changes — after brainstorm use /vp-crystallize / /vp-evolve → /vp-auto depending on the stage. Explicit override — state it in chat.</implementation_routing_guard>

Supports:

• Create a new session
• Continue a previous session
• Review a past session
• Landing page layout discovery (follow-up questions to finalize layout)
• In-session research (research during the brainstorm session on demand)
• UI Direction mode: create/update HTML prototype + notes under .viepilot/ui-direction/{session-id}/ — supports multi-page (pages/{slug}.html + hub index.html) and the ## Pages inventory hook in notes.md when pages/ exists (FEAT-007)
• Phase assignment (ENH-030): in every session, features/capabilities are assigned directly to Phase 1, Phase 2, Phase 3... — no MVP/Post-MVP/Future tiers. Session file stores assignments in the ## Phases section.
• Project meta intake (FEAT-009): after scope locked, before Completed / /end, if .viepilot/META.md (viepilot_profile_id) is missing — run sequential Q&A with proposals; read/write ~/.viepilot/profile-map.md; create ~/.viepilot/profiles/<slug>.md + binding per docs/dev/global-profiles.md. If a profile is already bound — skip intake by default (ask if change needed).
• UX walkthrough (FEAT-010 + ENH-019 + ENH-020): in --ui mode, proactively offered via AUQ when user mentions "research", "test UX", "walkthrough" — or when ≥1 UI signals accumulated. Runs 3 phases: simulates end-user (with content stress pass + stress recipes by archetype → Stress findings) → UX designer + web research → update index.html / pages/*.html / style.css and write ## UX walkthrough log in notes.md (sync hub + Pages inventory for multi-page).
• Background UI extraction (ENH-026 + ENH-060): automatically detects UI signal keywords in every brainstorm session (no --ui flag required). Auto-suggests itself (ENH-060): if the initial message contains ≥1 UI keyword, shows a proactive 🎨 banner immediately — mirrors Architect Design Mode's 🏗️ proactive activation. Accumulation starts at ≥1 signal (was ≥3); surfaces for confirmation when topic ends, user signals done, or ≥2 unique signals accumulated (was ≥5) — does not interrupt the main conversation.
• Idea → Architecture breakdown loop (ENH-061): structured 8-step flow from free idea collection → scope lock → Feature → Coverage mapping (maps each Phase 1 feature to an architect page + UI screen, outputs ## Coverage matrix in notes.md) → Architect Design → arch_to_ui_sync reverse sync (surfaces UI implications of architectural decisions; proactively offered via AUQ after arch updates) → UI Direction → completeness gate (CHECK 4: warns if any Phase 1 feature has no coverage in either mode, non-blocking). See Recommended Breakdown Ordering section in workflows/brainstorm.md.
• Unified workspace mode selection (BUG-018): after scope lock, before any design workspace is created, an AUQ prompt offers: Both / Architect only / UI Direction only / Neither. Architect auto-activate heuristic is deferred until after this selection; suppressed if user selects "Neither" or "UI Direction only".
• Admin & Governance coverage (ENH-063): Topic 6 in brainstorm template; proactive 🔐 heuristic fires on admin/governance keywords; admin coverage gate before /save for multi-user/SaaS/compliance projects; admin.html added to Architect workspace when applicable; notes.md ## admin YAML exported via crystallize.
• Content Management coverage (ENH-065): Topic 7 in brainstorm template; proactive 🗂️ heuristic fires on content-type/media/SEO keywords; content coverage gate before /save for projects with content layer; content.html added to Architect workspace; notes.md ## content YAML exported via crystallize.
• Admin Entity Management coverage (ENH-068): Topic 7 in brainstorm template; proactive 🗄️ heuristic fires on CRUD/entity/admin panel keywords; entity management coverage gate before /save for projects with DB entities (cross-references ERD); entity-mgmt.html added to Architect workspace; notes.md ## entity_mgmt YAML exported via crystallize.
• User Data Management coverage (ENH-066): Topic 9 in brainstorm template; proactive 👤 heuristic fires on user-account/privacy/auth keywords; user data coverage gate before /save for B2C/SaaS/GDPR projects; user-data.html added to Architect workspace; notes.md ## user_data YAML exported via crystallize.
• Architect Design Mode (FEAT-011): /vp-brainstorm --architect or auto-activate when ≥3 components/services detected; generate HTML workspace (architecture, data-flow, decisions, tech-stack, tech-notes, feature-map) with Mermaid diagrams; incremental update per decision; architecture review offered via AUQ when user asks to review decisions; machine-readable notes.md YAML schema.
• ERD page (ENH-027): Architect workspace includes erd.html — Mermaid erDiagram, entity list table, relationship summary; triggered by DB/entity/table/relationship keywords; notes.md ## erd YAML section exported to ARCHITECTURE.md ## Database Schema via crystallize Step 1D.
• User Use Cases page (ENH-028): Architect workspace includes user-use-cases.html — actor/use-case diagram (Mermaid flowchart), use case table; triggered by user/role/actor/story keywords; notes.md ## use_cases YAML section exported to PROJECT-CONTEXT.md ## User Stories & Use Cases via crystallize Step 1D.
• C4Context/Sequence/Deployment/APIs pages (ENH-029, 12-page workspace): Architect workspace expanded to 12 pages — sequence-diagram.html (per-scenario sequenceDiagram), deployment.html (infra graph + environments + CI/CD pipeline), apis.html (endpoint tables with HTTP method badges); page boundary rules table; trigger keywords for sequence/deploy/API; notes.md ## apis YAML section; deployment+APIs exported to ARCHITECTURE.md via crystallize Step 1D (sequence excluded — scenario docs are not architecture artifacts).
• Embedded Domain Mode (ENH-071): Activated when ≥2 embedded trigger keywords detected (MCU families: STM32/ESP32/nRF52/AVR/RISC-V/RP2040; concepts: firmware/bare-metal/GPIO/interrupt/HAL/bootloader/RTOS) OR --domain embedded flag. Adds 6 Architect workspace pages for embedded hardware artifacts, injects domain-specific topic probes, suppresses web-UI false-positives for hardware display keywords, and offers a firmware phase ordering template. crystallize exports 8 YAML sections to ARCHITECTURE.md hardware sections.

Creates/Updates:

• docs/brainstorm/session-{YYYY-MM-DD}.md

After: Ready for /vp-crystallize

Language configuration (ENH-032):

• Step 0 reads ~/.viepilot/config.json → BRAINSTORM_LANG = language.document (default: en).
• BRAINSTORM_LANG is used for brainstorm file storage and generated content.
• User session language takes precedence over config if different.
• Configure via: vp-tools config set language.document vi

Workflow version stamps (ENH-067):

• Session files record workflow_version (current ViePilot semver) at create/save time.
• upgrade_supplement_version stamped after gap-detection supplement completes (idempotency guard).

Upgrade gap detection (ENH-067):

• On --continue, compares session workflow_version vs. current Topics Template to detect missing topics.
• Proactive 🔄 upgrade banner (AUQ) lists missing topics; user can discuss inline, defer to /save, or skip.
• Inline supplement appended as ## Upgrade supplement (vX → vY) section; upgrade_supplement_version stamped for idempotency.
• After supplement: suggests /vp-crystallize --upgrade to patch project artifacts.</objective>

Reference URL Input (ENH-092)

When user provides a URL during brainstorm (competitor app, documentation, reference product):

• research-agent is dispatched with op: browse_url using vercel-labs/agent-browser
• Extracted insights added as ## Reference Research section in session output
• Covers: features, UX patterns, pricing model, tech stack clues
• Available as context for crystallize step
• Falls back to WebFetch if agent-browser not installed (static HTML only — warns if page is JS-rendered)

Examples:

/vp-brainstorm I want to build a task manager. Reference: https://linear.app
/vp-brainstorm --new Compare these two tools: https://notion.so https://coda.io

Multiple URLs: dispatches compare_products op → side-by-side comparison table. </context>

Embedded Domain Mode (ENH-071)

Activation: Automatically when ≥2 embedded keywords detected, or via --domain embedded flag. One-time 🔌 Embedded Domain Mode activated banner shown.

Topic probes injected when embedded_domain: true:

• MCU/Toolchain (Gap 2): MCU family, toolchain (GCC-ARM/Keil/IAR), build system (CMake/PlatformIO/West), debug interface (SWD/JTAG), flasher, SDK/HAL
• RTOS/Scheduling (Gap 3): bare-metal vs RTOS choice, task list, ISR table, resource protection strategy
• Power Budget (Gap 7): supply type, sleep modes, current targets, battery life (triggered by battery/power/sleep keywords)
• Safety/Compliance (Gap 10): IEC 61508/ISO 26262/DO-178C, watchdog, fault handlers (triggered by safety/SIL/ASIL keywords)
• Firmware Phase Template (Gap 9): Board Bring-Up → Drivers → RTOS → Middleware → Application → Integration Test → OTA

6 new Architect workspace pages (in .viepilot/architect/{session-id}/):

Pages linked in index.html under an Embedded nav section.

UI Direction false-positive suppression (Gap 6):

• LCD / OLED / TFT / display / screen in hardware context → routed to hw-topology peripherals, NOT UI Direction buffer
• Hardware context confirmed by: GPIO / SPI / I2C / driver / framebuffer / ILI9341 / SSD1306 / LVGL / u8g2 keywords
• 🎨 UI Direction Mode? banner suppressed when all display signals have hardware context

notes.md YAML sections written: ## hw_topology, ## pin_map, ## memory_layout, ## protocols, ## rtos_config, ## embedded_toolchain, ## power_budget, ## safety_config

crystallize Step 1D item 13 exports:

• ## Hardware Architecture (from hw_topology)
• ## Hardware Interface (from pin_map)
• ## Memory Map (from memory_layout)
• ## Communication Protocols (from protocols) — distinct from ## APIs
• ## RTOS & Task Model (from rtos_config)
• ## Toolchain & Build System (from embedded_toolchain)
• ## Power Budget (from power_budget)
• ## Safety & Reliability (from safety_config)
• ## Embedded Domain: true + MCU family written to PROJECT-CONTEXT.md
• UI Coverage Gate skipped; Hardware Coverage Check (driver-task completeness, non-blocking) applied instead

Key steps:

1. Detect existing sessions
2. Ask user intent (new/continue/review)
3. Load context if continuing
4. Run interactive Q&A with topic-based structure
5. If topic is a landing page: ask follow-up layout questions + reference 21st.dev to suggest sections/components
6. If topic needs UI/UX: create/update UI Direction artifacts under .viepilot/ui-direction/{session-id}/ — legacy: index.html + style.css + notes.md; multi-page: add pages/*.html, index.html as hub, and after each page change update ## Pages inventory in notes.md (see docs/user/features/ui-direction.md) 6b. When user types /research-ui or /research ui during a UI session: follow workflows/brainstorm.md exactly (FEAT-010) — do not merge into the short regular research step
7. If user requests research or needs to clarify a decision: research inline in the session and return to the topic
8. When a topic adds/modifies a capability: update ## Phases in the session per workflows/brainstorm.md
9. Before completing the session: step 5 — Project meta intake (FEAT-009) in workflows/brainstorm.md when binding is missing; sequential Q&A + profile-map disambiguation + write global profile + .viepilot/META.md
10. Save session with structured format (including ## Project meta intake (FEAT-009), research notes + UI direction references + ## Phases)
11. Suggest next action: /vp-crystallize

Stakeholder Generation (ENH-098): At session end (scope locked), automatically infers project stakeholders from the domain and creates .claude/agents/{project-slug}-{role}.md agent files. These are consumed by /vp-crystallize Step 1G for parallel context review. Skip with --no-stakeholders. </process>

Adapter Compatibility

AskUserQuestion Tool (ENH-048)

This skill uses adapter-aware interactive prompts. Behavior depends on your adapter:

Claude Code (terminal) — AUQ preload required (ENH-059): Before the first interactive prompt, call ToolSearch with query: "select:AskUserQuestion" to load the deferred tool schema. Only after ToolSearch succeeds can AskUserQuestion be invoked. If ToolSearch returns an error, fall back to plain-text numbered list for that session.

When AskUserQuestion is not available on other adapters, the skill automatically falls back to plain-text numbered list prompts — no configuration required.

Prompts using AskUserQuestion in this skill:

• Session intent (continue / review / new — Step 2)
• Landing page layout selection (Step 4 — Layout A/B/C/D)
• Mid-session structured choices (≥2 discrete named options — BUG-022): topic direction, section priority, angle selection, and any numbered decision with enumerable options
• Session-transition/next-steps prompt (session-flow choices: save/crystallize, update UI artifacts, continue discussing — BUG-023)

Skill Registry Integration (FEAT-020)

When UI Direction Mode is active, vp-brainstorm automatically:

1. Loads ~/.viepilot/skill-registry.json (written by vp-tools scan-skills)
2. Matches installed skills by capabilities to current UI signals
3. Silently applies matched skill best_practices to HTML generation
4. Records matched skills in notes.md ## skills_used

No prompt is shown — integration is transparent. Skills with relevant capabilities (e.g., ui-generation, component-design, responsive-layout) are detected automatically.

Skill decisions are locked at crystallize time (see /vp-crystallize). /vp-auto executes with those locked decisions — no re-asking.

Install skills via vp-tools install-skill <source>. See: docs/user/features/skill-registry.md