diff --git a/.grok/skills/grok-chat-model-map/README.md b/.grok/skills/grok-chat-model-map/README.md
new file mode 100644
index 00000000..64af64c3
--- /dev/null
+++ b/.grok/skills/grok-chat-model-map/README.md
@@ -0,0 +1,45 @@
+# grok-chat-model-map
+
+Grok skill: map **Chat** modes (Auto / Fast / Expert / Heavy / Build) to the **Grok 4.6** generation and keep Chat ≠ API ≠ Build ≠ Imagine.
+
+## Install (Grok Build / grok.com skills)
+
+```bash
+# From this repo (plan branch or after merge):
+mkdir -p ~/.grok/skills
+cp -R .grok/skills/grok-chat-model-map ~/.grok/skills/
+
+# Or unzip a release pack into ~/.grok/skills/grok-chat-model-map/
+```
+
+Then in Grok:
+
+```text
+ACTIVATE GROK_CHAT_MODEL_MAP
+```
+
+or `/grok-chat-model-map`
+
+## Layout
+
+```
+grok-chat-model-map/
+├── SKILL.md
+├── README.md
+└── references/
+    ├── CHEAT_SHEET.md
+    ├── mode-catalog.md
+    └── PASTE_PACK.md
+```
+
+## Verify
+
+1. Open https://grok.com Chat picker — expect Auto · Fast · Expert · Heavy · Build  
+2. Activate skill → “Grok Chat model map locked…” + pick card  
+3. Confirm it never recommends sending `Auto` as an API `model`
+
+## Links
+
+- PR: https://github.com/FineComputer14451/Grok-Imagine-Cinematic-Studio/pull/48  
+- Docs: https://docs.x.ai/developers/models · https://docs.x.ai/developers/grok-4-6  
+- Sibling: `.grok/skills/imagine-model-overrides`
diff --git a/.grok/skills/grok-chat-model-map/SKILL.md b/.grok/skills/grok-chat-model-map/SKILL.md
new file mode 100644
index 00000000..c000012a
--- /dev/null
+++ b/.grok/skills/grok-chat-model-map/SKILL.md
@@ -0,0 +1,119 @@
+---
+name: grok-chat-model-map
+description: >-
+  Map Grok Chat modes (Auto / Fast / Expert / Heavy / Build) to the Grok 4.6
+  generation and keep Chat vs API vs Build vs Imagine surfaces straight. Use
+  when the user asks which Chat mode to pick, confuses modes with API model
+  ids, or mixes Chat with Imagine/Build. Activate with ACTIVATE GROK_CHAT_MODEL_MAP
+  or /grok-chat-model-map.
+when-to-use: >-
+  ACTIVATE GROK_CHAT_MODEL_MAP; Grok Chat mode picker; Auto vs Fast vs Expert;
+  Heavy mode; Build from Chat; grok-4.6 chat mapping; do not send Auto as API model
+argument-hint: "[auto|fast|expert|heavy|build|tree|api]"
+user-invocable: true
+metadata:
+  author: FineComputer14451
+  short-description: Map Grok Chat modes to Grok 4.6 (not API ids)
+  version: "1.0.0"
+  verified-picker: "2026-09-06"
+---
+
+# Grok Chat Model Map
+
+Canonical **Chat mode** skill for [grok.com](https://grok.com) Chat (also iOS / Android / X Grok).
+
+**Verified picker (2026-09-06 ET):** `Auto` · `Fast` · `Expert` · `Heavy` · `Build`
+
+Begin every activation with: **"Grok Chat model map locked…"**  
+Then emit the pick card (Mode / Why / Not).
+
+Supporting files:
+- `references/CHEAT_SHEET.md` — paste-ready user card
+- `references/mode-catalog.md` — modes + API ids + aliases
+- `references/PASTE_PACK.md` — short share paste blocks
+- `README.md` — install into `~/.grok/skills`
+
+## Hard rules
+
+1. **Never invent picker labels.** Only: Auto, Fast, Expert, Heavy, Build.
+2. **Never send Chat mode names as API `model` strings.** `Auto` / `Expert` ≠ `grok-4.6`.
+3. **Grok 4.6** is the Chat/Code generation on [docs.x.ai/models](https://docs.x.ai/developers/models) (“Chat: Grok 4.6”) — **not** a fifth depth tile in the verified Chat picker.
+4. **Build** in the Chat picker opens **Build Mode** (in-chat apps → `*.grok.me`), **not** the terminal Grok Build CLI (`grok`). Distinct products; see [xAI](https://x.ai/news/grok-build-mode).
+5. **Imagine** stills/video → skill `imagine-model-overrides` (sidebar Imagine), not this skill.
+6. **Cinematic Studio** `tools/models.py` is a different registry (do not merge).
+
+## Mode catalog (Chat UI)
+
+| Mode | Kind | When to use | Relation to Grok 4.6 |
+|------|------|-------------|----------------------|
+| **Auto** | Router | Everyday default | Routes Fast-like vs Expert-like; not an API id |
+| **Fast** | Latency path | Quick facts, short asks | Same Chat product family; lighter route — API id not shown in UI |
+| **Expert** | Deep reasoning | Hard analysis, long reasoning | Best match for flagship Chat on **4.6** generation |
+| **Heavy** | Multi-agent max | Hardest multi-step depth | Separate Heavy stack (not “slow Expert”) |
+| **Build** | Build Mode (product) | Shareable sites/apps/games/dashboards in chat | Opens **Build Mode** → preview / `grok.me` — **not** Termux `grok` CLI |
+
+Full tables: `references/mode-catalog.md`.
+
+## Argument shortcuts
+
+| Arg | Action |
+|-----|--------|
+| `auto` / `fast` / `expert` / `heavy` / `build` | Lock that mode + why |
+| `tree` | Print surface decision tree only |
+| `api` | Print API id reminder (never mode names as `model`) |
+| _(empty)_ | Ask goal in one line, then recommend |
+
+## Quick picks
+
+| You want… | Pick |
+|-----------|------|
+| Default / mixed day | **Auto** |
+| Speed | **Fast** |
+| Depth / hard reasoning | **Expert** |
+| Max multi-agent depth | **Heavy** |
+| Shareable in-chat app / site | **Build** (Build Mode) |
+| Repo / lab coding (CLI) | Terminal **Grok Build** (`grok`) — not this tile |
+| Stills / video | **Imagine** (other skill) |
+| HTTP API call | API id e.g. `grok-4.6` — not a Chat mode |
+
+## Surface tree
+
+```
+Chat thread               → Auto | Fast | Expert | Heavy
+Chat picker → Build       → Build Mode (in-chat preview → *.grok.me)
+Sidebar Imagine           → Imagine image/video models
+api.x.ai / SDK            → grok-4.6, grok-4.5, …
+Terminal Grok Build CLI   → grok / grokhunter (repo edits; default grok-4.6)
+Studio registry           → tools/models.py (separate)
+```
+
+## Output card (always)
+
+```markdown
+## Grok Chat pick
+- Mode: …
+- Why: …
+- Not: API id / Imagine / Studio registry
+```
+
+## Tier note
+
+Verified session 2026-09-06: all five modes visible with **no lock badges**. Quotas can still differ by plan (Heavy often needs SuperGrok Heavy for full compute).
+
+
+## Build Mode vs Grok Build CLI
+
+| | Chat **Build** (Build Mode) | Terminal **Grok Build** |
+|--|-----------------------------|-------------------------|
+| Where | grok.com / iOS / Android Chat | Termux / desktop `grok` |
+| Output | Live preview + `*.grok.me` (or custom domain) | File edits, shell, agents |
+| Use for | Prototypes you share as a link | Repo / GrokHunter lab work |
+
+Never treat Chat **Build** as “open my phone rootfs.”
+Official: https://x.ai/news/grok-build-mode
+
+## References
+
+- Plan / pack PR: https://github.com/FineComputer14451/Grok-Imagine-Cinematic-Studio/pull/48
+- Docs: https://docs.x.ai/developers/models · https://docs.x.ai/developers/grok-4-6
+- Sibling skill: `imagine-model-overrides` (Imagine Quality/Fast · Video 1.0/1.5)
diff --git a/.grok/skills/grok-chat-model-map/grok-chat-model-map-skill.tar.gz b/.grok/skills/grok-chat-model-map/grok-chat-model-map-skill.tar.gz
new file mode 100644
index 00000000..2ef54098
Binary files /dev/null and b/.grok/skills/grok-chat-model-map/grok-chat-model-map-skill.tar.gz differ
diff --git a/.grok/skills/grok-chat-model-map/references/CHEAT_SHEET.md b/.grok/skills/grok-chat-model-map/references/CHEAT_SHEET.md
new file mode 100644
index 00000000..1a66ca4e
--- /dev/null
+++ b/.grok/skills/grok-chat-model-map/references/CHEAT_SHEET.md
@@ -0,0 +1,55 @@
+# Grok Chat Mode Cheat Sheet
+**For https://grok.com Chat — Auto / Fast / Expert / Heavy / Build**
+
+Verified picker **2026-09-06**: Auto · Fast · Expert · Heavy · Build  
+Skill: `ACTIVATE GROK_CHAT_MODEL_MAP`  
+Plan PR: https://github.com/FineComputer14451/Grok-Imagine-Cinematic-Studio/pull/48  
+Install: copy folder → `~/.grok/skills/grok-chat-model-map/`  
+Models docs: https://docs.x.ai/developers/models · Grok 4.6: https://docs.x.ai/developers/grok-4-6
+
+---
+
+## 30-second pick
+
+| You want… | Click in Chat picker |
+|-----------|----------------------|
+| Default / mixed | **Auto** |
+| Speed | **Fast** |
+| Hard reasoning / depth | **Expert** |
+| Max multi-agent depth | **Heavy** |
+| Shareable in-chat app/site | **Build** (Build Mode) |
+| Repo / lab coding | Terminal `grok` (not Chat Build) |
+| Stills / video | Sidebar **Imagine** (not this picker) |
+
+**Grok 4.6** = current Chat/Code generation (docs). It is **not** a separate fifth depth tile in the verified Chat picker.
+
+---
+
+## Mode → meaning
+
+| Mode | Kind | Notes |
+|------|------|-------|
+| Auto | Router | Picks Fast-like or Expert-like path — **not** an API model string |
+| Fast | Latency | Light / quick answers |
+| Expert | Deep | Best Chat match for **4.6**-generation flagship reasoning |
+| Heavy | Multi-agent | Separate Heavy path — not “Expert but slower” |
+| Build | Build Mode | In-chat apps → `*.grok.me` — **not** Termux CLI |
+
+---
+
+## Do not confuse
+
+| Surface | What you pick |
+|---------|----------------|
+| grok.com **Chat** | Auto / Fast / Expert / Heavy (/ Build) |
+| **api.x.ai** | `grok-4.6`, `grok-4.5`, … — never send `Auto` or `Expert` as `model` |
+| **Grok Build** TUI (`grok`) | Repo agent on device; default ≈ `grok-4.6` — **≠** Chat Build Mode |
+| **Build Mode** (Chat → Build) | Preview + publish `*.grok.me` ([news](https://x.ai/news/grok-build-mode)) |
+| **Imagine** | Quality/Fast stills · Video 1.0/1.5 → `imagine-model-overrides` |
+| Cinematic Studio | `tools/models.py` registry (different map) |
+
+---
+
+## Tier note
+
+On the verified 2026-09-06 grok.com session, all five modes were **visible with no lock badges**. Quotas can still differ by plan (Heavy often needs SuperGrok Heavy for full compute).
diff --git a/.grok/skills/grok-chat-model-map/references/PASTE_PACK.md b/.grok/skills/grok-chat-model-map/references/PASTE_PACK.md
new file mode 100644
index 00000000..6bb53cc7
--- /dev/null
+++ b/.grok/skills/grok-chat-model-map/references/PASTE_PACK.md
@@ -0,0 +1,34 @@
+# Paste packs — Grok Chat model map
+
+## Short (chat / X)
+
+```
+Grok Chat picker = modes, not API ids.
+Auto=router · Fast=speed · Expert=depth (Grok 4.6 gen) · Heavy=multi-agent · Build=coding agent.
+Never send Auto/Expert as api.x.ai model — use grok-4.6.
+ACTIVATE GROK_CHAT_MODEL_MAP · PR #48
+```
+
+## Medium (share / forum)
+
+```
+Grok Chat picker = modes, not API ids (verified 2026-09-06).
+Auto=router · Fast=speed · Expert=depth (Grok 4.6 gen) · Heavy=multi-agent · Build=coding agent.
+Never send Auto/Expert as api.x.ai model — use grok-4.6.
+Stills/video → Imagine sidebar. ACTIVATE GROK_CHAT_MODEL_MAP · PR #48
+```
+
+## Activation line (in Grok Chat)
+
+```
+ACTIVATE GROK_CHAT_MODEL_MAP
+```
+
+Optional args: `expert` · `fast` · `heavy` · `build` · `tree` · `api`
+
+
+## Build Mode ≠ Grok Build CLI
+
+Chat picker **Build** = Build Mode (sites/apps in chat → grok.me).
+Terminal **Grok Build** = `grok` / `grokhunter` on your machine (repo edits).
+Same word, two products. https://x.ai/news/grok-build-mode
diff --git a/.grok/skills/grok-chat-model-map/references/mode-catalog.md b/.grok/skills/grok-chat-model-map/references/mode-catalog.md
new file mode 100644
index 00000000..7cd95de6
--- /dev/null
+++ b/.grok/skills/grok-chat-model-map/references/mode-catalog.md
@@ -0,0 +1,72 @@
+# Grok Chat mode catalog
+
+Source: live grok.com Chat picker **2026-09-06** + https://docs.x.ai/developers/models + https://docs.x.ai/developers/grok-4-6
+
+## Chat UI modes (canonical)
+
+| Mode | Kind | Typical use | Notes |
+|------|------|-------------|-------|
+| `Auto` | Router | Default | Not an API model string |
+| `Fast` | Latency | Short / simple | API id not exposed in UI |
+| `Expert` | Deep reasoning | Hard tasks | Best Chat match for **Grok 4.6** generation |
+| `Heavy` | Multi-agent | Max depth | Separate Heavy path |
+| `Build` | Build Mode | Shareable apps in chat | Opens Build Mode → `*.grok.me` — **not** CLI `grok` |
+
+### Aliases (spoken → mode)
+
+```
+default, router, automatic     → Auto
+quick, speed, light, low-latency → Fast
+deep, think, reasoning, smart  → Expert
+max, multi-agent, team         → Heavy
+prototype, website, app, game, grok.me → Build (Build Mode)
+code, coding, agent, cli, tui, repo     → Terminal Grok Build (`grok`) — not Chat Build
+```
+
+Do **not** invent extra picker tiles (no “Grok 4.6” depth mode in verified snapshot).
+
+## API / developer ids (not the Chat dropdown)
+
+| API id | Role vs Chat |
+|--------|----------------|
+| `grok-4.6` | Recommended Chat + Code on API; Expert/Auto sit on this generation |
+| `grok-4.5` | Prior flagship |
+| `grok-4.3` | Long-context / alternate |
+| `grok-4.20-0309-reasoning` | Reasoning variant |
+| `grok-4.20-0309-non-reasoning` | Non-reasoning / faster analogue |
+| `grok-4.20-multi-agent-0309` | Multi-agent analogue (Heavy-adjacent family) |
+| `grok-build-0.1` | Build/coding agent model — not a Chat depth mode |
+
+### Forbidden as API `model`
+
+```
+Auto, Fast, Expert, Heavy, Build
+auto, fast, expert, heavy, build
+```
+
+## grok.com UI mapping
+
+| Control | Meaning |
+|---------|---------|
+| Chat picker → Auto/Fast/Expert/Heavy | Chat reasoning / routing modes |
+| Chat picker → Build | **Build Mode** (in-chat create/publish) — not Termux CLI |
+| Sidebar → Imagine | Image/video models (see imagine-model-overrides) |
+
+## Cross-skill
+
+| Need | Skill |
+|------|-------|
+| Chat mode pick | **grok-chat-model-map** (this) |
+| Imagine Quality/Fast · Video 1.0/1.5 | `imagine-model-overrides` |
+| Studio ROLE_DEFAULTS | Cinematic Studio `tools/models.py` |
+
+
+## Build Mode vs Grok Build CLI
+
+| | Chat picker **Build** | Terminal **Grok Build** (`grok`) |
+|--|----------------------|----------------------------------|
+| Product | Build Mode ([news](https://x.ai/news/grok-build-mode)) | Coding agent CLI |
+| Where | grok.com / iOS / Android | Local machine / Termux |
+| Output | Preview + `*.grok.me` | Repo edits / shell / agents |
+
+API id `grok-build-0.1` is a **model id**, not the Chat Build tile.
diff --git a/docs/superpowers/plans/2026-09-06-grok-chat-46-model-mapping.md b/docs/superpowers/plans/2026-09-06-grok-chat-46-model-mapping.md
new file mode 100644
index 00000000..adc22db8
--- /dev/null
+++ b/docs/superpowers/plans/2026-09-06-grok-chat-46-model-mapping.md
@@ -0,0 +1,215 @@
+# Grok Chat 4.6 Model Mapping Implementation Plan
+
+> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
+
+**Goal:** Document and keep current a faithful map of **Grok Chat** (grok.com / iOS / Android / X Grok) consumer model **modes** and how they relate to **Grok 4.6** and other xAI backends — not the Cinematic Studio registry.
+
+**Architecture:** Treat Chat UI labels (Auto / Fast / Expert / Heavy / Build) as the product surface. Map each label to the best-known routing behavior and API-adjacent model family. Separate **Chat modes** from **API model IDs** (`grok-4.6`, etc.). Prefer live picker verification over third-party blogs.
+
+**Tech Stack:** Markdown plan + optional cheat-sheet skill for grok.com; verify against live grok.com picker and https://docs.x.ai/developers/models.
+
+## Global Constraints
+
+- Surface: **Grok Chat** only (grok.com, iOS Grok, Android Grok, X-integrated Grok) — not Cinematic Studio `tools/models.py`, not GrokHunter V9 host pickers (except as optional cross-links)
+- Flagship chat intelligence (Aug 2026+): **Grok 4.6** per xAI (“Chat: Grok 4.6” on docs.x.ai/models)
+- Consumer picker uses **modes** (Auto/Fast/Expert/Heavy) more than raw API slugs
+- Tier gates matter: free vs SuperGrok / Premium+ vs SuperGrok Heavy
+- Never invent live picker labels — verify on grok.com when logged in
+- Do not print account secrets or subscription receipts
+
+---
+
+## Working map (baseline — verify live)
+
+### A. Chat UI modes (consumer)
+
+**Verified live on grok.com: 2026-09-06 (ET)** — picker shows exactly:
+
+`Auto` · `Fast` · `Expert` · `Heavy` · `Build`
+
+| UI mode | Behavior | Likely routing | Notes |
+|---------|----------|----------------|-------|
+| **Auto** | Default router by query complexity | Chooses Fast-like or Expert-like path | Do not assume a single API slug |
+| **Fast** | Low-latency, light answers | Non-/low-reasoning path | Good for facts, short asks |
+| **Expert** | Forced deeper reasoning | Flagship reasoning stack (4.6-era Expert) | Super / Premium+ often required for full access |
+| **Heavy** | Multi-agent / max depth | Heavy / multi-agent stack | SuperGrok Heavy |
+| **Build** | Build Mode (in-chat create/publish) | Opens Build Mode → preview / `*.grok.me` ([news](https://x.ai/news/grok-build-mode)) | **Not** the terminal Grok Build CLI; distinct from Auto/Fast/Expert/Heavy |
+
+No separate “Grok 4.6” named pin was visible in this picker snapshot — 4.6 is the generation behind Chat/Expert (per docs.x.ai), not a fifth depth mode in this UI.
+
+### A2. Mode → Grok 4.6 generation (relationship)
+
+Sources: live picker 2026-09-06 · https://docs.x.ai/developers/models (“Chat: Grok 4.6”) · https://docs.x.ai/developers/grok-4-6
+
+| Mode | Kind | Relation to Grok 4.6 |
+|------|------|----------------------|
+| **Auto** | **Router** (not a model id) | Picks a Fast-like or Expert-like path per query; does not expose `grok-4.6` as a slug |
+| **Fast** | Latency / light path | Same Chat product generation family; may use a lighter or non-/low-reasoning route — **exact API id not exposed in UI** |
+| **Expert** | Forced deep reasoning | Best match for flagship Chat intelligence on the **4.6 generation** (docs: Chat → Grok 4.6) |
+| **Heavy** | Multi-agent / max depth | Separate product path (Heavy stack); not “Expert but slower” |
+| **Build** | Build Mode product | Opens **Build Mode** (in-chat apps → `*.grok.me`). Not a Chat reasoning depth. Terminal **Grok Build** CLI is a separate product (default coding model **grok-4.6**) |
+
+**Rule of thumb:** In Grok Chat, you pick a **mode**. On the API, you pick an **id** like `grok-4.6`. Do not treat `Auto`/`Fast`/`Expert`/`Heavy` as API model strings.
+
+### A3. Tier matrix (access)
+
+| Mode | Verified on this grok.com session (2026-09-06) | Typical gate (product docs / common knowledge — confirm on account) |
+|------|-----------------------------------------------|---------------------------------------------------------------------|
+| Auto | Available (selected) | Broad / default |
+| Fast | Available (no lock badge in screenshot) | Broad |
+| Expert | Available (no lock badge in screenshot) | Often SuperGrok / X Premium+ for full quotas |
+| Heavy | Available (no lock badge in screenshot) | Commonly SuperGrok Heavy for full Heavy compute |
+| Build | Available (no lock badge in screenshot) | Build entitlement / plan-dependent |
+
+**Observed:** On the verified account snapshot, all five options appeared selectable with **no lock/upgrade badges** in the open picker. Tier **quotas** and rate limits can still differ even when the label is unlocked — do not equate “visible” with “unlimited.”
+
+**TBD (do not invent):** Exact free-tier vs paid locks for other accounts/regions; iOS/Android parity; X-in-app picker differences.
+
+### A4. Surface decision tree (stop the PR #47 mix-up)
+
+```
+Where are you working?
+│
+├─ grok.com / iOS / Android / X → Chat thread
+│     → Use Chat modes: Auto | Fast | Expert | Heavy
+│     → Build in that picker → Build Mode (in-chat → *.grok.me; ≠ terminal grok CLI)
+│     → Sidebar Imagine → Imagine image/video models (not Chat modes)
+│
+├─ api.x.ai / console / SDK
+│     → Use API ids: grok-4.6, grok-4.5, grok-4.3, grok-4.20-*, grok-build-0.1, …
+│     → Do NOT send "Auto" / "Expert" as model strings
+│
+├─ Grok Build TUI / CLI (`grok`, host ~/.grok/config.toml)
+│     → Use Build catalog + optional /model aliases (chat-expert, multi, auto)
+│     → Default coding intelligence: grok-4.6
+│
+└─ Cinematic Studio / Imagine Cinematic registry
+      → tools/models.py ROLE_DEFAULTS (separate product map; closed PR #47)
+```
+
+**Quick picks**
+- Everyday chat → **Auto**
+- Need speed → **Fast**
+- Need depth → **Expert** (4.6-generation Chat)
+- Need multi-agent max → **Heavy**
+- Need shareable in-chat app → Chat **Build** (Build Mode)
+- Need repo/lab coding → terminal **Grok Build** CLI (`grok`), not Chat Build / not Expert
+- Need stills/video → **Imagine**, not Chat modes
+
+### B. API / developer IDs (docs.x.ai — not the Chat dropdown)
+
+| API id | Role vs Chat |
+|--------|----------------|
+| `grok-4.6` | Recommended for Chat + Code on API; Chat UI Expert/Auto often sit on this generation |
+| `grok-4.5` | Prior flagship; may still appear in API / migrations |
+| `grok-4.3` | Long-context / alternate stack |
+| `grok-4.20-0309-reasoning` | Reasoning variant |
+| `grok-4.20-0309-non-reasoning` | Non-reasoning / faster path analogue |
+| `grok-4.20-multi-agent-0309` | Multi-agent analogue (Heavy-adjacent family) |
+| `grok-build-0.1` | Build/coding agent model — **not** Chat mode |
+
+### C. Explicit non-goals (wrong surface)
+
+- Cinematic Studio `STACK_CONTRACT` / Imagine Image-Video ROLE_DEFAULTS → see closed PR #47
+- Grok Build `/model chat-expert|multi|auto` host TOML aliases → GrokHunter / Build, not grok.com Chat
+- Imagine Quality vs Fast stills → Imagine surface (Selector teammate / grok.com/imagine)
+
+---
+
+## File map
+
+| File | Responsibility |
+|------|----------------|
+| `docs/superpowers/plans/2026-09-06-grok-chat-46-model-mapping.md` | This plan |
+| Optional later: `docs/guides/GROK_CHAT_MODELS.md` or skill pack | User-facing cheat sheet for Chat modes |
+| Optional later: community / share cheat sheet | Same map, paste-ready |
+
+---
+
+## Task 1: Live-verify Chat picker labels
+
+**Goal:** Replace third-party mode lists with a dated screenshot-backed inventory from grok.com.
+
+- [x] Open grok.com Chat model/mode picker (browser verify 2026-09-06)
+- [x] List every label exactly: Auto, Fast, Expert, Heavy, Build
+- [x] Note tier badges / locked items — none shown on verified 2026-09-06 snapshot (all five selectable); quotas may still differ
+- [x] Capture screenshot (attached in PR / plan assets)
+- [x] Update the Working map table with **Verified (2026-09-06)**
+- [x] Commit: `docs(chat): verify grok.com Chat mode picker labels`
+
+---
+
+
+### Build Mode vs Grok Build CLI (naming)
+
+xAI **Build Mode** (Chat picker → Build): create websites/apps/games in chat and publish to `*.grok.me` — Early Beta, [announcement](https://x.ai/news/grok-build-mode).
+
+**Grok Build** CLI (`grok`): terminal coding agent for repos (GrokHunter lab uses this).
+
+Do not document Chat Build as “hop to Termux grok” without this split.
+
+## Task 2: Mode → Grok 4.6 relationship note
+
+**Goal:** One clear paragraph + table: which modes are “on the 4.6 generation,” which are routers, which are Heavy-only.
+
+- [x] Write: Auto = router (not a model id)
+- [x] Write: Expert ≈ flagship reasoning on current Chat generation (**4.6** as of Aug 2026 docs)
+- [x] Write: Fast ≈ latency path (API id not exposed in UI — marked uncertain)
+- [x] Write: Heavy = separate multi-agent product path; Build = surface hop
+- [x] Cross-link docs.x.ai models + grok-4.6 pages
+- [x] Commit: `docs(chat): map Auto/Fast/Expert/Heavy/Build to Grok 4.6 generation`
+
+---
+
+## Task 3: Tier matrix
+
+**Goal:** Who can pick what.
+
+- [x] Document verified-session availability (all five visible, no lock badges)
+- [x] Note typical gates without inventing prices
+- [x] Mark free-tier / other-platform parity as TBD
+- [x] Commit: `docs(chat): Chat mode access by subscription tier`
+
+---
+
+## Task 4: Chat vs API vs Build vs Imagine — decision tree
+
+**Goal:** Stop future mix-ups (the PR #47 failure mode).
+
+- [x] One-page decision tree added (section A4)
+- [x] Commit: `docs(chat): surface decision tree Chat vs API vs Build vs Imagine`
+
+---
+
+## Task 5: Optional cheat sheet skill / community pack
+
+**Goal:** Paste-ready card for grok.com skills / sharing if the user wants distribution.
+
+- [x] Skill pack: `.grok/skills/grok-chat-model-map/` (SKILL.md, README, CHEAT_SHEET, mode-catalog, PASTE_PACK)
+- [x] Mirror verified Auto/Fast/Expert/Heavy/Build table; no Studio registry content
+- [x] Commit on plan branch / PR #48
+
+---
+
+## Task 6: Close the loop on PR #47
+
+**Goal:** History is clear.
+
+- [x] Close PR #47 with comment that Studio plan was wrong target
+- [x] Link this Chat plan PR from #47 comments / PR #48 body references #47
+- [ ] Do **not** merge Studio mapping plan into main unless user later asks
+
+---
+
+## Done when
+
+1. Live picker labels are listed with a verification date (or auth-block is documented)
+2. Auto / Fast / Expert / Heavy vs Grok 4.6 relationship is explicit and non-contradictory with docs.x.ai
+3. Chat vs API vs Build vs Imagine surfaces cannot be confused in the plan
+4. PR #47 remains closed as the wrong-surface artifact
+
+## Out of scope
+
+- Changing Cinematic Studio `tools/models.py`
+- Rewriting GrokHunter `install_v9_grok_models.sh`
+- Imagine model overrides (owned by Imagine Selector / PR #45 track)
diff --git a/docs/superpowers/plans/assets/2026-09-06-grok-chat-picker.png b/docs/superpowers/plans/assets/2026-09-06-grok-chat-picker.png
new file mode 100644
index 00000000..ed2929d0
Binary files /dev/null and b/docs/superpowers/plans/assets/2026-09-06-grok-chat-picker.png differ