@easysociable/cli 0.1.18 → 0.1.20

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@easysociable/cli",
-  "version": "0.1.18",
+  "version": "0.1.20",
   "description": "Agent-first CLI for EasySociable: Content Engine, Template, Slideshow, and MCP",
   "type": "module",
   "license": "MIT",
@@ -8,11 +8,10 @@
   "homepage": "https://easysociable.com",
   "repository": {
     "type": "git",
-    "url": "git+https://github.com/murphywuwu/8MRR-Project.git",
-    "directory": "apps/cli"
+    "url": "git+https://github.com/murphywuwu/easysociable-mcp.git"
   },
   "bugs": {
-    "url": "https://github.com/murphywuwu/8MRR-Project/issues"
+    "url": "https://github.com/murphywuwu/easysociable-mcp/issues"
   },
   "keywords": [
     "easysociable",

package/skills/ONBOARDING.md

@@ -19,14 +19,13 @@ V1 expects an Agent (Grok Bot, Codex, Claude Code, …) to drive setup. The webs
 
    Default skills:
 
-   - `easysociable-slideshow` — turns a draft into slideshow pages you preview and download in the dashboard
-   - `easysociable-template` — creates page templates
+   - `easysociable-slideshow` — turns a topic into carousel pages you edit and download in Studio
 
    Optional Content Engine skill (install via CLI only — not in the public
    content-engine git repo; operations still follow that engine’s `CLAUDE.md`):
 
    ```bash
-   easysociable install --skills content,template,slideshow --yes
+   easysociable install --skills content,slideshow --yes
    ```
 
    Get a Content Engine core (preferred):

package/skills/easysociable-slideshow/SKILL.md

@@ -3,8 +3,8 @@ name: easysociable-slideshow
 description: >
   Turn a topic into an editable, ready-to-post carousel (Slideshow Run) in
   EasySociable Studio. Research facts, pick a proven hook and structure
-  (Formula), adapt pages from one Deck Style's catalog layouts, render-check
-  every page locally, then create the Run and hand the user a Studio link.
+  (Formula), lay pages out from reference designs (the user's, a Formula's,
+  or the catalog's, shuffled), render-check every page locally, then create the Run and hand the user a Studio link.
   Ask the user only for what only they know (their Brand, once; their own
   images/screenshots). Final pixels come only from Studio Download.
 ---
@@ -40,12 +40,14 @@ Ask only for what only the user knows:
 - None → ask in **one** message, everything skippable:
   "What name and @handle should appear on your carousels? (Optional: an
   avatar/logo image path.) Reply 'skip' to post without a signature."
-  Create it with `easysociable brands create --file brand.json` using the
-  palette of the chosen Deck Style (`references/DECK_STYLES.md`).
-  Skipped → no Brand: header shows the topic, the CTA has no handle, and the
-  Deck Style palette goes in `theme.json` as
-  `{ "color": { "canvas", "text", "textMuted", "accent", "primary" } }`
-  (without a Brand or theme Studio falls back to its dark default).
+  Create it with `easysociable brands create --file brand.json`
+  (`{ "name", "handle", "colors", "logoAssetId" }`) using the default palette
+  that matches the deck's style (`references/REFERENCES.md`). An avatar: upload
+  it first (step 5's `images upload`) and pass the returned `asset.id` as
+  `logoAssetId`; `brands get` then returns its `logoUrl`.
+  Skipped → no Brand: header shows the topic, the CTA has no handle, and that
+  palette goes in `theme.json` (without a Brand or theme Studio falls back to
+  its dark default).
 
 ### 2. Facts
 
@@ -67,8 +69,12 @@ niche). Use a card when it fits; otherwise write your own. Never force one.
   `mistake-list`, a single piece of news or a claim → no list recipe.
 - `easysociable hooks recommend --json` → one mold per hook type (list-tease,
   contrast, number, identity, …), ranked by evidence.
-- `easysociable formulas recommend --recipe <recipe> --platform <p> --fact-count <n> --json`
-  when the shape is a recipe.
+- **Always** look for a structure first:
+  `easysociable formulas recommend [--recipe <recipe>] --platform <p> --fact-count <n> --json`
+  (`--recipe` when the shape is a list recipe; without it you see every
+  structure and judge fit yourself). Near-equal cards come back shuffled.
+  When the user asked for something different from last time, add
+  `--exclude <formulaId>` for the Formula that Run used.
 - Use a card only if `stop` is false **and** it fits this content (a news
   explainer is not an "N things" list). Prefer `published`; a `hypothesis`
   card is thin evidence: use it only when it clearly fits.
@@ -83,19 +89,38 @@ niche). Use a card when it fits; otherwise write your own. Never force one.
 - One idea per page, ≤ 45 words per page. Page 1 is the hook. Last page asks
   for one action (comment, save or follow).
 
-### 4. Style and pages
-
-Pick one Deck Style (`references/DECK_STYLES.md`): the Brand's style if it has
-one, else by content. Styles that need images only when you have the images.
-
-For each page, take the family that fits its purpose and adapt its source:
-
-```bash
-easysociable templates source --family <familyId> --platform <platform> --json
-```
-
-Follow "Adapting a family" in `DECK_STYLES.md`. Put the result in the pack as a
-generated page:
+### 4. Look and pages
+
+Structure (step 3) and look are chosen separately. The look comes from a
+**reference**; the first rule that applies wins:
+
+1. The user names a Run or says "same as last time / like this one" →
+   "Reuse a Run" below.
+2. The user shares a post (screenshot or link) → study its grid, type scale,
+   background, density and decoration; take the closest catalog references as
+   structure and restyle them to match the post. Never copy its words or images.
+3. A Formula with a Look → the Look's designs (Formula handoff below).
+4. The user asks for a different look ("换个样子", "something else") → first
+   write 3–5 art directions that differ from each other and from this Brand's
+   previous Run (one line each: grid, type treatment, background, density,
+   decoration). Drop any the previous Run used, take the least conventional
+   one, and pick references whose `style` fits it.
+5. Otherwise → the content decides: get candidates for the cover with
+   `easysociable templates references --role hook,cover --json`
+   (results are shuffled) and take the first that fits the content. Its
+   `style` is the deck's style; for each other page ask for that page's roles
+   with `--style <style>`. A number → `metric,proof`; a list item →
+   `point,value,steps`; a comparison → `compare`; a real product →
+   `screenshot`. No match in that style → any style, keeping the cover's chrome.
+
+Don't filter out references that need an image just because you have none:
+most covers in the library have an image area. Use one and fill that area
+without a photo (references/REFERENCES.md, "A reference's image area"). Keep a
+placeholder only where the content needs a real screenshot or photo (step 5).
+`--no-image` is for when you specifically want designs drawn with type alone.
+Read each chosen design with `easysociable templates source --family <familyId> --json`
+and adapt it as in `references/REFERENCES.md` ("Adapting a reference" and
+"Avoid"). Put the result in the pack as a generated page:
 
 ```json
 { "role": "hook", "slots": { "text_0": "…", "text_1": "…" },
@@ -104,8 +129,10 @@ generated page:
 ```
 
 `slots` maps each `data-variable` to its plain text. Generated pages carry no
-`layoutFamilyId` / `pageTemplateId`. A Formula with a Look uses the Look's
-families as the sources instead of a Deck Style.
+`layoutFamilyId` / `pageTemplateId`.
+
+In the final message, add one line: "Want future carousels to look like this?
+Say 'same as this'. Want another look? Say 'different look'."
 
 ### 5. Images
 
@@ -118,7 +145,7 @@ families as the sources instead of a Deck Style.
   placeholder slot** (as many pages as need one), never a text-only layout. Size it into the composition (usually the
   lower half of the page) and label it with exactly what belongs there, e.g.
   "Add: Grok Bot app screenshot". The user selects it in Studio and clicks
-  the upload button to replace it. Markup and SVG in `DECK_STYLES.md`
+  the upload button to replace it. Markup and SVG in `references/REFERENCES.md`
   ("Image placeholder").
 - An image you do have: upload it
   (`easysociable images upload --file body.json` with
@@ -188,8 +215,40 @@ different topic, or when the user asks for an alternative version to compare.
 order and roles); `title_template` with `{X}` = the topic is the arc. Write copy
 from each `beat.fill` and your facts; never copy the formula's example posts.
 With `Look: L-…` (else the formula's first `looks` entry) the page source is
-`look.slides["<n>"]` or `look.layouts[beat.role]`: adapt that family exactly as
-in step 4. Then continue at step 5.
+`look.slides["<n>"]` or `look.layouts[beat.role]` (a design's `familyId`):
+read it with `templates source --family` and adapt it as in step 4. The Look
+wins over rule 5 of step 4 but not over a Run or post the user named.
+Then continue at step 5.
+
+## Reuse a Run (prompt names a `run_…` for a new topic)
+
+A new topic → a **new** Run; the source Run is never changed. Never ask which
+mode; decide from the words and the content:
+
+- **Template** — "using run_x as the template", "same as last time", "like
+  this one" (last time = the newest Run of this Brand, `slideshows list`).
+  Keep every page's look: only slot content changes.
+- **Style** — "in the style / feel / vibe of run_x", or Template was asked but
+  the new content does not fit its page roles (a "N tips" Run for an "A vs B"
+  topic). Take its colours, type treatment, chrome and spacing; lay pages out
+  fresh in step 4.
+
+Template steps:
+
+1. `easysociable slideshows get --id <runId> --json`. Its pages' `role`,
+   `layout` and `slots` are the source (they include the user's Studio edits).
+2. Write facts and the arc for the new topic (steps 2–3, no Formula search).
+3. Pages by role: one-off roles (hook, cta) copy that page's `layout`
+   verbatim; repeated roles (point, step) reuse the source page of that role
+   **as many times as the new content needs** (4 tips → 4 pages, 6 → 6). Drop
+   roles the new content has none of.
+4. Change only text inside `data-variable` elements and send matching
+   `slots`. Too long → rewrite shorter; never resize boxes or move elements.
+   Every source image becomes a placeholder for the new topic (step 5);
+   Brand avatar/logo stay.
+5. Step 6 check, then create (step 7). In the reply, say in one line what
+   differs from the source ("6 tips, so 6 point pages instead of 4") and that
+   the user can say "back to 4 pages" or "different layout".
 
 ## Rules
 
@@ -201,7 +260,7 @@ in step 4. Then continue at step 5.
 3. Revisions edit the same Run page by page (step 8): layout, text and images
    can all change. Page count and order are fixed for now (no add, delete,
    reorder) — if the user needs that, say so; create a new Run only when they agree.
-4. Free: up to 8 pages and 3 Downloads per month; paid: 20 pages, unmetered.
+4. Every plan: up to 20 pages. Free: 3 Downloads per month and 1 Brand; Creator: unlimited, 3 Brands; Studio: unlimited, 25 Brands. `BRAND_LIMIT_REACHED` → tell the user their plan's limit and the suggested action; never create a Run without the Brand silently.
 5. Identity: create with the user's own login, not `workspace_local` /
    `test-key` (unless they asked for `--allow-local-fixture`). Studio says
    "belongs to another account" → recreate with the current login.
@@ -220,6 +279,6 @@ in step 4. Then continue at step 5.
 
 ## Handoff boundary
 
-The Content Engine owns knowledge pages and Formula research. The Template
-skill owns authoring new catalog layouts. This skill turns a topic into a Run
+The Content Engine owns knowledge pages and Formula research. Reference
+designs are authored internally (Template skill); this skill only reads them. This skill turns a topic into a Run
 and opens Studio.

package/skills/easysociable-slideshow/references/CONTENT_PACK.md

@@ -1,9 +1,8 @@
 # ContentPack v1
 
-A page is either **generated** (`layout: { studioJson, canvas }` adapted from a
-Deck Style family — the default, see `DECK_STYLES.md`) or a **catalog page**
-(`layoutFamilyId` / `pageTemplateId`, used as-is; only for Formula Looks the
-user asked to copy exactly). `role` is a weak tag from the closed enum, not a
+A page is **generated**: `layout: { studioJson, canvas }` adapted from a
+reference design (see `REFERENCES.md`). Catalog pages (`layoutFamilyId` /
+`pageTemplateId`) are legacy; do not create new Runs with them. `role` is a weak tag from the closed enum, not a
 sequence position. `slots` are public values only. Do not add a
 `deck_template_id`, `bindingMap`, AST, ContentDocument, or product `arc` /
 `pageSequence`.

package/skills/easysociable-slideshow/references/REFERENCES.md

@@ -0,0 +1,104 @@
+# Reference designs
+
+A reference is a published catalog design (one canvas: the source post's) made
+from a real post. Pages are **adapted from references**: you keep a reference's
+composition, type scale, spacing and chrome, and replace content, colours
+(Brand) and canvas size. You never copy one verbatim and never draw a page with
+no reference at all.
+
+Find references for a page role, shuffled, as short summaries:
+
+```bash
+easysociable templates references --role hook,cover [--style <style>] [--exclude <familyIds>] [--no-image] --json
+easysociable templates source --family <familyId> --json   # the design's Studio source
+```
+
+Roles overlap in the catalog, so pass related ones together: `hook,cover` ·
+`point,value,steps,checklist` · `metric,proof,evidence` · `compare` ·
+`screenshot` (needs an image) · `recap` · `cta,follow`. Styles: `minimal`,
+`editorial`, `swiss`, `dark`, `bold_promo`.
+
+## Adapting a reference (the rules that make it look designed)
+
+1. Start from the reference's source. Keep its chrome (header, pill, page
+   number, rule, footer) only if the deck keeps it on every page; keep its
+   margins (`left` of the first column), type scale and fonts.
+2. Fit the Run's canvas: a reference made for another canvas (3:4, 9:16, 4:5)
+   keeps its left margin and type scale; re-space vertically, never squash text.
+3. Change text, then fix each text box: explicit `<br />` line breaks, and
+   `height` = lines × lineHeight + ~12% of fontSize for descenders.
+4. Remove what you do not need; add only elements copied from another reference
+   with the same `style` (rule rows, a stat row).
+4a. **A reference's image area** when you have no image and the content does
+   not need a real one: keep the area's box and fill it from the content with a
+   large number or keyword, a short quote with an oversized quote mark, an
+   accent-colour block carrying one word, or one icon from
+   `easysociable icons search` on a tinted block. Never a grey placeholder or an
+   empty box on the cover. If nothing fits, delete the area and give the
+   headline its space.
+5. Brand chrome: the header label is the Brand name (or the topic when there is
+   no Brand); the CTA names the Brand handle. Never leave a reference's original
+   name or handle (e.g. "Will.AI.m") on a page. **Bind them** so a Brand change
+   updates every page: the element holding only the Brand name gets
+   `data-brand="name"`, only the handle `data-brand="handle"`, and an avatar /
+   logo `<img>` gets `data-brand="avatar"` with the Brand's `logoUrl` as `src`.
+   Without a Brand, leave them unbound (the topic is not the Brand).
+6. Accent colour only on large text (≥ 36px, or ≥ 28px bold) and blocks. Small
+   labels use the text or muted colour.
+7. Semantic `data-type` values (`title`, `subtitle`, `kicker`, `metric`,
+   `caption`) get Brand colours forced by Studio. Text on an accent or image
+   block uses `label`, `body` or `statement` with an explicit colour.
+8. One deck = one `style`. No reference twice in a Run unless the page purpose
+   repeats (several numbered points), and no two adjacent pages with the same
+   composition.
+
+## Avoid (what makes every Run look the same)
+
+- The same cover composition as this Brand's previous Run, unless the user
+  asked for "same as last time".
+- Header label + pill + page number + "SWIPE →" on every page by habit: chrome
+  comes from the chosen reference, or there is none.
+- An accent-coloured second half of the headline on more than two pages.
+- A deck that is only left-aligned text stacks: at least one page's focal point
+  is not a headline (a big number, an image, a quote, a comparison).
+- A generic "Save this / Follow for more" CTA when the content suggests a
+  specific action (comment a word, try the prompt, read the source).
+
+## Default palettes (no Brand)
+
+Use one in `theme.json` as `{ "color": { "canvas", "text", "textMuted", "accent", "primary" } }`
+(`primary` = `accent`), matched to the deck's style:
+
+| Style | canvas / text / textMuted / accent |
+|---|---|
+| `editorial` | `#F4EEE7` / `#121A2B` / `#4A4F5C` / `#D4450F` |
+| `minimal`, `swiss` | `#FFFFFF` / `#15191E` / `#384352` / `#445164` |
+| `dark` | `#06080C` / `#FFFFFF` / `#C7CBD3` / `#4C9BFF` |
+| `bold_promo` | `#F9F9F9` / `#0A0A0A` / `#555555` / `#0066DD` |
+
+## Image placeholder
+
+A replaceable image slot for a real screenshot or photo the user will add.
+Studio shows an upload button when it is selected.
+
+```jsx
+<img data-type="image" data-variable="image_0" data-image-mode="slot_default"
+  alt="Add: <what goes here>" src="data:image/svg+xml;base64,<SVG>"
+  style={{ position: 'absolute', left: 88, top: 760, width: 904, height: 388, objectFit: 'cover', borderRadius: 24 }} />
+```
+
+`<SVG>` is base64 of a grey box with an image icon and the label, sized to the slot:
+
+```xml
+<svg xmlns='http://www.w3.org/2000/svg' width='904' height='388' viewBox='0 0 904 388'>
+  <rect width='904' height='388' fill='#DDD6CC'/>
+  <g transform='translate(416,124)' fill='none' stroke='#8C8781' stroke-width='5' stroke-linejoin='round'>
+    <rect width='72' height='56' rx='8'/><path d='M6 50 L26 28 L40 42 L50 34 L66 50'/><circle cx='52' cy='16' r='6'/>
+  </g>
+  <text x='50%' y='230' text-anchor='middle' font-family='Inter, Arial, sans-serif' font-size='32' font-weight='600' fill='#6F6A64'>Add: Grok Bot app screenshot</text>
+</svg>
+```
+
+Put the same data URL in the page's `slots` under `image_N`. Number slots
+`image_0`, `image_1`, … per page. Use the style's own colours for the box
+(canvas-adjacent grey on light styles, `#1A1F27` on dark ones).

package/skills/easysociable-slideshow/workflows/CREATE_RUN.md

@@ -1,7 +1,7 @@
 # Create Run
 
 Follow the Workflow in `SKILL.md` (login → Brand → facts → hook/structure →
-Deck Style pages → images → `scripts/check-pages.mjs` → create).
+look + pages from references → images → `scripts/check-pages.mjs` → create).
 
 Notes for updates:
 

package/skills/easysociable-template/SKILL.md

@@ -1,11 +1,11 @@
 ---
 name: easysociable-template
 description: >
-  Create single-page social image templates for EasySociable via the unified
-  Template MCP or `easysociable` CLI. Default create path is a Layout Family:
-  one master canvas plus derived multi-platform variants (e.g. XHS 3:4, IG/Threads 4:5,
-  TikTok 9:16, LinkedIn 1:1) with hand-tuned layouts per preset. Use when the user wants
-  to author a page template from a reference image or text brief, write a shared visual
+  Internal tool (opt-in install): turn a real post's page (reference image) or a
+  text brief into a single-canvas reference design for the slideshow skill's
+  reference library, then bind it to a Formula Look. One canvas only (the
+  source post's size, e.g. TikTok 9:16); no multi-platform variants. Use when the
+  maintainer wants to author a reference from a reference image or text brief, write a shared visual
   design philosophy, explore layout concepts, co-create Theme/preview tokens, pull icons
   via search_icons/get_icon, open Authoring Studio, or publish. Do not use for
   fill-slots hosted PNG jobs (`create_render_job` is retired), multi-slide slideshow
@@ -21,7 +21,7 @@ server. Do not install or invoke `easysociable-template`, `easysociable-slidesho
 
 **This skill owns template authoring:**
 
-1. **Author** templates as a **Layout Family** (master + multi-platform variants) → Studio + Publish (`workflows/MULTI_PLATFORM_FAMILY.md`).
+1. **Author** reference designs: **one canvas**, the source post's size → Studio + Publish, then bind to a Formula Look (`workflows/BIND_FORMULA_LOOK.md`). Other platforms are re-laid out by the slideshow agent at generation time, so never derive variants.
 2. Fill-slots Hosted Render (`create_render_job`) is **retired** as a paid delivery path. Customer PNG packs come from Slideshow Studio Download.
 
 It is a template factory. It is **not** a multi-slide slideshow compositor and it is **not** a PNG API.
@@ -41,9 +41,9 @@ Out of scope (hand off):
 2. Submit only public Gateway DTOs. Never invent or request private Safe AST / Binding Map / RenderPlan.
 3. Do not start Local Runtime, read content library SQLite, or call Content MCP/CLI for Template work.
 4. Catalog and create targets are **single-page** only. Never treat site Carousel packs as Agent templates.
-5. After `create` / `createDraft`, the primary authoring path is `authoringStudioUrl` (`/authoring/{draftId}`). **Default:** run `workflows/MULTI_PLATFORM_FAMILY.md` (derive + hand-tune variants) before handoff unless the user asked for one platform only. Always send the handoff in `AUTHORING_STUDIO.md` (edit → **Save** → tell Agent to publish). Do not require a Hosted Preview Job for interactive authoring.
+5. After `create` / `createDraft`, the primary authoring path is `authoringStudioUrl` (`/authoring/{draftId}`). One canvas only (never `families derive`). Always send the handoff in `AUTHORING_STUDIO.md` (edit → **Save** → tell Agent to publish). Do not require a Hosted Preview Job for interactive authoring.
 6. Studio has **Save only** — no Publish button. Publish only after the user saves and **explicitly** asks to publish; then run `workflows/PUBLISH_FROM_DRAFT.md`.
-7. **Do not Publish without complete P0 selection metadata** (`CATALOG_METADATA.md`). Gateway enforces this. Each family variant carries its own `platforms` / `canvasPreset` / `canvas`.
+7. **Do not Publish without complete P0 selection metadata** (`CATALOG_METADATA.md`). Gateway enforces this. What the slideshow agent filters on must be right: `supportedRoles`, `contentArchetypes`, `requiresImage`, `visualStyle`, `bestFor`.
 8. Theme/preview: co-create `ThemeTokens` on draft create (`theme.create`) → Family `preview_tokens` (ADR-0033). No authoring TemplateTheme row. Colors/fonts via semantic `--es-*` on page nodes.
 9. Layout model is **absolute positioning** on a **flat canvas tree** (see prior authoring rules / `AUTHORING_SOURCE.md`).
 10. **Runtime layout CSS:** put exact geometry in JSX inline `style={{...}}`. Never depend on Tailwind arbitrary-value classes for geometry.
@@ -76,12 +76,12 @@ easysociable auth status --json
 
 | User intent | Workflow |
 |---|---|
-| Text description → template (incl. multi-platform) | `DESIGN_BRIEF.md` → … → `AUTHORING_STUDIO.md` → **`MULTI_PLATFORM_FAMILY.md`** (derive + hand-tune) |
-| Reference image → template | `AUTHORING_FROM_IMAGE.md` → … → `AUTHORING_STUDIO.md` → **`MULTI_PLATFORM_FAMILY.md`** + reference-compare QA on master |
-| “只要某一个平台” | Same authoring path; skip derive / only named presets in `MULTI_PLATFORM_FAMILY.md` |
+| Text description → reference | `DESIGN_BRIEF.md` → … → `AUTHORING_STUDIO.md` → publish |
+| Reference image (a real post's page) → reference | `AUTHORING_FROM_IMAGE.md` → … → `AUTHORING_STUDIO.md` (reference-compare QA) → publish |
+| A viral post → its Formula's Look | each page via `AUTHORING_FROM_IMAGE.md` → publish → **`BIND_FORMULA_LOOK.md`** |
 | Need gradient / photo surface / icons only | `SURFACE_REVIEW.md` |
 | **Add / change an icon** (arrow, chevron, check, …) | **MUST** `SURFACE_REVIEW.md` § Icons — search → confirm → get_icon. **Do not** hand-draw SVG paths. |
-| Already have authoring payload / continue Studio | `workflows/AUTHORING_STUDIO.md` (+ family derive if variants still missing) |
+| Already have authoring payload / continue Studio | `workflows/AUTHORING_STUDIO.md` |
 | User says publish / 发布该模板 (after Save) | `workflows/PUBLISH_FROM_DRAFT.md` |
 | Browse published single-page templates | `list` / `get_template` + public metadata |
 | **Fill a published template → PNG/JPEG** | `workflows/FILL_AND_RENDER.md` |
@@ -89,14 +89,14 @@ easysociable auth status --json
 
 ## Core model
 
-- `workflows/MULTI_PLATFORM_FAMILY.md` — **default** multi-platform create (Family + derive + hand-tune)
+- `workflows/BIND_FORMULA_LOOK.md` — published references → a Formula's Look (role → design)
 - `references/PLATFORM_SPECS.md`
 - `references/CATALOG_METADATA.md`
 - `references/SLOTS.md` — public slot contract for fill-slots render
 - `references/AUTHORING_SOURCE.md`
 - `references/THEME_TOKENS.md` — Family `preview_tokens` (no TemplateTheme)
 - `references/EXTERNAL_RESOURCES.md`
-- `references/GATEWAY.md` — CLI / MCP including families derive + `render`
+- `references/GATEWAY.md` — CLI / MCP
 - `workflows/FILL_AND_RENDER.md` — get_template → slots → Hosted Render
 - Authoring V2.1 workflows + validators as before
 

package/skills/easysociable-template/references/GATEWAY.md

@@ -12,15 +12,13 @@ easysociable templates update-draft --id <id> --file ./patch.json --json
 easysociable templates validate --id <id> --json
 easysociable templates publish --id <id> --json
 easysociable templates families get --id <familyId> --json
-easysociable templates families derive --id <familyId> --presets 4:5,9:16,1:1 --json
-easysociable templates families migrate-legacy --json
 easysociable icons search --query "arrow" --json
 easysociable icons get --id <id> --idempotency-key <key> --json
 easysociable renders create --template-id <id> --slots ./slots.json --format png --idempotency-key render_1 --json
 easysociable renders wait --id <id> --json
 ```
 
-Multi-platform create flow (derive + hand-tune): `workflows/MULTI_PLATFORM_FAMILY.md`.
+One canvas per reference; binding to a Formula: `workflows/BIND_FORMULA_LOOK.md`.
 
 All requests go through Gateway. The client receives public summaries, slot
 contracts, stable Image references and render Manifests only. It never receives

package/skills/easysociable-template/references/PLATFORM_SPECS.md

@@ -120,20 +120,18 @@ No platform chrome assumptions. Still pick a concrete preset (`1:1`, `4:5`, `3:4
 
 ---
 
-## Layout Family (multi-platform create)
+## One canvas per reference
 
-Creating a template **defaults** to a Layout Family with one master page + derived presets — see `workflows/MULTI_PLATFORM_FAMILY.md`.
-
-Same-preset platforms share one variant tree (**Instagram + Threads → `4:5`**). Do not invent a second 4:5 page for Threads.
-
-Typical vertical bucket when master is `3:4` (Xiaohongshu): also derive `4:5`, `9:16`, `1:1`. Absolute layout cannot share one AST across those presets — hand-tune each variant after derive.
+A reference design has exactly one canvas: the source post's (a TikTok post → `9:16`).
+Never derive other presets; the slideshow agent re-lays a reference out for the
+Run's platform.
 
 ## Selection rules for Agents
 
-1. If user names one platform only → set master `platforms: [that]` + that platform’s **Recommended** row; skip family derive unless they later ask for more.  
-2. If user names several **or** does not limit platforms → pick a master Recommended canvas, then derive the default vertical targets (`MULTI_PLATFORM_FAMILY.md`).  
+1. Canvas = the source post's platform **Recommended** row (text brief with no post → TikTok 9:16).  
+2. One canvas only; no `families derive`.  
 3. Never mix presets across slides of one deck without an explicit user choice (IG/Threads/TikTok crop to first slide).  
-4. Each family variant’s P0 `selectionMetadata` must list only platforms that match that page’s `canvasPreset`.  
+4. P0 `selectionMetadata` lists only platforms that match the page's `canvasPreset`.  
 5. Specs above are **practical authoring targets** (2026 guides + Meta help for IG). Re-verify if a platform announces breaking crop changes.
 
 ## Sources (research snapshot)

package/skills/easysociable-template/workflows/AUTHORING_FROM_BRIEF.md

@@ -4,12 +4,12 @@ Primary intake when the user describes a single-page template in words (no refer
 
 ## Goal
 
-One publishable **single-page** template **Family** (master canvas + multi-platform variants by default): Design Brief → Design Philosophy → 2–3 Concepts → Asset Plan → Theme lock → master layout → Authoring Studio → `MULTI_PLATFORM_FAMILY.md`.
+One publishable **single-page, single-canvas** reference design: Design Brief → Design Philosophy → 2–3 Concepts → Asset Plan → Theme lock → layout → Authoring Studio → publish.
 
 ## Steps
 
-1. Confirm this is **one page**, not a slideshow. If they want a multi-slide deck, stop and hand off (`OUT_OF_SCOPE.md`). Multi-platform variants of the same page are expected — that is not a slideshow.
-2. Read `references/PLATFORM_SPECS.md`. Lock **master** platform + recommended canvas, and the default vertical derive targets unless the user limited platforms (`MULTI_PLATFORM_FAMILY.md`).
+1. Confirm this is **one page**, not a slideshow. If they want a multi-slide deck, stop and hand off (`OUT_OF_SCOPE.md`).
+2. Read `references/PLATFORM_SPECS.md`. Lock platform + recommended canvas (one canvas).
 3. **Mandatory — V2.1 design plan first** (no reference image):
    - Follow `workflows/DESIGN_BRIEF.md`
    - Follow `workflows/DESIGN_PHILOSOPHY.md`; lock shared `philosophy` (+ companion `design-philosophy.md`)
@@ -30,7 +30,7 @@ One publishable **single-page** template **Family** (master canvas + multi-platf
    - `authoringSource` per `references/AUTHORING_SOURCE.md` (flat tree + text-nesting exception)
    - Theme per `references/THEME_TOKENS.md` — use `themeTokens` from validated `theme-tokens.json` for `theme.create`
 5. Run `scripts/validate-authoring-source.mjs <authoring-source.json>` and fix every error.
-6. Continue with `workflows/AUTHORING_STUDIO.md` (`source.kind = text_brief`) → **`MULTI_PLATFORM_FAMILY.md`** (derive + hand-tune + QA each variant) before handoff.
+6. Continue with `workflows/AUTHORING_STUDIO.md` (`source.kind = text_brief`) → Visual QA before handoff.
 
 ## Checklist pages
 

package/skills/easysociable-template/workflows/AUTHORING_FROM_IMAGE.md

@@ -79,7 +79,7 @@ If unsure whether a region is “illustration” vs “icon”: prefer **image s
 9. Propose Theme reuse or `ThemeTokens` create from `theme-tokens.json` (`references/THEME_TOKENS.md`).
 10. Show the user: geometry summary, **image inventory → slot map**, selection metadata, Theme. Call out any region you could not match and why. Proceed when they accept.
 11. Run `scripts/validate-authoring-source.mjs <authoring-source.json>` and fix every error.
-12. Continue with `workflows/AUTHORING_STUDIO.md` (`source.kind = image_reference`, brief = path or short note) → **`MULTI_PLATFORM_FAMILY.md`** (derive + hand-tune other presets). Run Visual QA **including reference compare on the master** before handoff; still QA each derived variant for fit/clipping.
+12. Continue with `workflows/AUTHORING_STUDIO.md` (`source.kind = image_reference`, brief = path or short note) Run Visual QA **including reference compare** before handoff. One canvas only; after publish, bind it with `BIND_FORMULA_LOOK.md` when it comes from a Formula's post.
 
 ## Limits (state upfront when relevant)
 

package/skills/easysociable-template/workflows/AUTHORING_STUDIO.md

@@ -18,11 +18,11 @@ Complete intake first:
 ## Layout Family (multi-platform)
 
 New drafts create a **Layout Family** (`familyId`) plus a master canvas page.
-**Default create path:** after the master draft exists, follow `MULTI_PLATFORM_FAMILY.md` (derive + hand-tune variants) before handoff unless the user asked for one platform only.
+One canvas per reference: never derive platform variants.
 
 - Authoring URL: `/authoring/{draftId}?platform=…`. Top bar **platform chips** switch variants (do not resize one AST).
 - Preview skin = Family `preview_tokens` from `theme.create`; fill/render uses Brand snapshot first (ADR-0033).
-- Legacy drafts without `familyId` auto-wrap on `getDraft` / `listDrafts`. Bulk: `easysociable templates families migrate-legacy --json`.
+- Legacy drafts without `familyId` auto-wrap on `getDraft` / `listDrafts`.
 
 ## Flow
 
@@ -47,9 +47,9 @@ AST or Binding Maps.
 
 3. Read `data.id` / `data.templateId` / `data.familyId` / `data.authoringStudioUrl` (and `data.variants` when present).
 
-4. **Multi-platform (default):** run `workflows/MULTI_PLATFORM_FAMILY.md` — derive target presets, hand-tune each variant, QA each canvas. Skip only when the user asked for a single platform.
+4. One canvas only: do not run `families derive`.
 
-5. **Before handoff, run `workflows/VISUAL_QA.md`** on the master (and each variant from step 4). Open the actual Studio render; inspect screenshots, not only source. When a draft fails, fix authoring JSON, validate, update that page id, and re-render:
+5. **Before handoff, run `workflows/VISUAL_QA.md`** on the page. Open the actual Studio render; inspect screenshots, not only source. When a draft fails, fix authoring JSON, validate, update that page id, and re-render:
 
 ```bash
 node scripts/validate-authoring-source.mjs ./authoring-source.json

package/skills/easysociable-template/workflows/BIND_FORMULA_LOOK.md

@@ -0,0 +1,39 @@
+# Bind references to a Formula Look
+
+A Formula is the structure distilled from a real viral post; its **Look** says
+which reference design each slide uses. The slideshow skill reads it
+(`look.slides["<n>"]`, else `look.layouts[<role>]`) whenever that Formula is used.
+
+## When
+
+After publishing the references for a post's pages (one reference per distinct
+page layout, via `AUTHORING_FROM_IMAGE.md`).
+
+## Steps
+
+1. `easysociable formulas get --id <F-…> --json` — note its `beats` (slide order
+   and roles) and existing `looks`.
+2. Map each beat to a published design's `familyId`
+   (`easysociable templates list --canvas 9:16 --json` shows `familyId`):
+   - `layouts`: one design per role (`hook`, `point`, `cta`, …);
+   - `slides`: only for a slide whose design differs from its role's default
+     (1-based slide number → `familyId`).
+3. Write the Look:
+
+   ```json
+   { "id": "L-<short-name>",
+     "ref": { "platform": "tiktok", "platform_post_id": "<the post's id>" },
+     "layouts": { "hook": "fam_…", "point": "fam_…", "cta": "fam_…" },
+     "slides": { "2": "fam_…" } }
+   ```
+
+4. There is no public write API for Formulas: hand the JSON to the maintainer
+   to append to that Formula's `looks` in the Directus admin (a new Look goes
+   last, so the Formula's default Look — the first — does not change unless
+   they choose to reorder).
+5. Check: `easysociable formulas get --id <F-…> --json` shows the Look, and
+   every `familyId` in it resolves with
+   `easysociable templates source --family <familyId> --json`.
+
+Never bind an agent-authored design that was not made from a real post: the
+reference library only holds proven designs.

package/skills/easysociable-template/workflows/DESIGN_BRIEF.md

@@ -16,7 +16,7 @@ Skip this workflow when `AUTHORING_FROM_IMAGE.md` applies and a reference image
 
 ### 1. Gather intent (batch questions; skip known answers)
 
-- **Master** platform + canvas from `references/PLATFORM_SPECS.md`, plus default multi-platform derive targets (`MULTI_PLATFORM_FAMILY.md`) unless the user wants one platform only
+- Platform + canvas from `references/PLATFORM_SPECS.md` (one canvas: the source post's)
 - Page roles (1–2) + contentFamily / archetypes / density (`CATALOG_METADATA.md`)
 - Content job: what must the viewer understand in one glance?
 - Must-have slots: title / metric / image / CTA / fixed N checklist lines…
@@ -31,8 +31,7 @@ Use this structure (Chinese or English to match the user):
 
 ## 1. Job to be done
 - One-sentence purpose of this page
-- Master platform + canvas (preset + pixels) — cite PLATFORM_SPECS recommended row
-- Multi-platform targets (default vertical bucket unless user limited) — `MULTI_PLATFORM_FAMILY.md`
+- Platform + canvas (preset + pixels) — cite PLATFORM_SPECS recommended row
 
 ## 2. Visual constraints
 - Desired/forbidden moods or brand rules