@event4u/agent-config 14.9.0 → 14.11.0

package/dist/agent-src/skills/design-system-capture/SKILL.md

@@ -10,6 +10,9 @@ workspaces:
 packs:
   - frontend-design
 token_budget_class: rich
+scope:
+  write: []
+  verification_reason: "execution declares no handler, so this skill runs nothing of its own — every write is the calling agent's, under the rules that govern it. No command can prove a scope the skill never executes."
 execution:
   type: manual
 ---
@@ -68,8 +71,30 @@ should apply consistently across all UI tasks in this project. Unlike
 
 ## Design philosophy
 One sentence: [density/formality/complexity philosophy]
+
+## Owned components
+| Component | Status | Story file | Registry item |
+|---|---|---|---|
+| [Name] | stable \| experimental \| deprecated | [path to *.stories.tsx] | [registry item name, or —] |
 Example: "Information-dense dashboard; optimize for power users, not casual exploration."
 
+## Owned components — filled from the tree, never from memory
+
+The § Owned components table is populated by **globbing `*.stories.tsx`** (and, when a
+Storybook manifest is reachable, reading it — see
+[`storybook-workshop`](../storybook-workshop/SKILL.md) § MCP path). One row per component
+the project owns: its name, its status, its story file, and its registry item where the
+library publishes one.
+
+**Never write a row from memory.** A remembered component is exactly the row a reader will
+trust and a `ui-component-architect` run will then skip re-checking — an inventory that is
+wrong in the direction of "we already have this" causes the duplicate it exists to prevent.
+A component with no story file gets a row with an empty story cell, not an omission: absent
+from the table reads as "does not exist".
+
+`deprecated` rows stay in the table. Deleting one loses the only durable record that the
+component should not be reached for again.
+
 ## Visual system
 
 ### Elevation and radius
@@ -194,46 +219,16 @@ Example: "Dashboard = control panel for power users, not an exploration space."
 
 ## Importing an extracted design system
 
-A consumer can reverse-engineer an existing site/repo's look-and-feel with any
-external static-extraction tool and hand the result to this skill as a
-`design-system.json` artifact. We own the **import contract**, not the crawler:
-the package never ships the Playwright runtime, a font-bundler, or a `.skill`
-auto-installer (out of scope). Full schema is lazy-loaded from
-[`references/design-system-json.md`](references/design-system-json.md) — read it
-only when an import is requested.
-
-**The supply path is real, and it is one command.** Where an extractor's output
-is not already in this shape, run it through the three-lane adapter first —
-`/design-system:import <file>`, which accepts a native artifact, a DTCG token
-file, or an extraction tool's raw JSON and emits the contract. Lanes and
-documented producers:
-[`references/design-system-json.md`](references/design-system-json.md)
-§ Extractor compatibility. The adapter is offline and pure — it does not change
-what this skill owns.
-
-**Import procedure:**
-
-1. Read `design-system.json`; reject it if `source` (kind + ref + captured_at)
-   is missing — no provenance, no import.
-2. Diff every field against the current `DESIGN.md`.
-3. Surface a **per-field confirm/merge proposal** — the artifact is *observed,
-   not authoritative* (mirrors `source-discovery`). Never write silently.
-4. **Conflict with a registered brand value** (a confirmed `.tokens.json` /
-   brand token) → **flag, never auto-apply** (`brand-source-of-truth`:
-   consumer brand wins). Precedence: brand tokens > confirmed `DESIGN.md` >
-   imported observation.
-5. On the human's accept, persist the chosen fields into `DESIGN.md`. Where the
-   consumer wants a token source of truth, hand the mapped DTCG fields to
-   [`brand-to-tokens`](../brand-to-tokens/SKILL.md) / `design-tokens` to
-   materialise `.tokens.json` — do not invent a parallel token format.
-
-**Two sources, one shape:**
-
-- **External target** (a site/repo you don't own) → an external tool emits the
-  artifact; import it here.
-- **Current repo** → prefer [`existing-ui-audit`](../existing-ui-audit/SKILL.md);
-  it already inventories the codebase and can emit the same `design-system.json`
-  shape, so the import path is identical either way.
+A consumer can reverse-engineer an existing site/repo's look-and-feel with any external
+static-extraction tool and hand the result to this skill as a `design-system.json`
+artifact. We own the **import contract**, not the crawler. The artifact is *observed, not
+authoritative*: never write it in silently, and a conflict with a registered brand value is
+flagged rather than applied (`brand-source-of-truth`).
+
+Import contract, the `/design-system:import` supply path, the five-step procedure and the
+two source shapes — read only when an import is requested:
+[`references/import-procedure.md`](references/import-procedure.md). Full schema:
+[`references/design-system-json.md`](references/design-system-json.md).
 
 ## How the design skills consume these documents
 

package/dist/agent-src/skills/design-system-capture/references/import-procedure.md

@@ -0,0 +1,50 @@
+# Importing an extracted design system
+
+> The import contract, the supply path, the five-step procedure, and the two source shapes
+
+_Split out of [`SKILL.md`](../SKILL.md) § Importing an extracted design system, which keeps
+the one-line obligation and routes here. The split is ADR-217's rich-class ceiling doing its
+job: adding the owned-components inventory took the skill to 3,764 tokens against a 3,500
+ceiling, and this block is the largest self-contained one that a reader needs only when an
+import is actually requested — the same lazy-load posture the schema reference already has._
+
+A consumer can reverse-engineer an existing site/repo's look-and-feel with any
+external static-extraction tool and hand the result to this skill as a
+`design-system.json` artifact. We own the **import contract**, not the crawler:
+the package never ships the Playwright runtime, a font-bundler, or a `.skill`
+auto-installer (out of scope). Full schema is lazy-loaded from
+[`design-system-json.md`](design-system-json.md) — read it
+only when an import is requested.
+
+**The supply path is real, and it is one command.** Where an extractor's output
+is not already in this shape, run it through the three-lane adapter first —
+`/design-system:import <file>`, which accepts a native artifact, a DTCG token
+file, or an extraction tool's raw JSON and emits the contract. Lanes and
+documented producers:
+[`design-system-json.md`](design-system-json.md)
+§ Extractor compatibility. The adapter is offline and pure — it does not change
+what this skill owns.
+
+**Import procedure:**
+
+1. Read `design-system.json`; reject it if `source` (kind + ref + captured_at)
+   is missing — no provenance, no import.
+2. Diff every field against the current `DESIGN.md`.
+3. Surface a **per-field confirm/merge proposal** — the artifact is *observed,
+   not authoritative* (mirrors `source-discovery`). Never write silently.
+4. **Conflict with a registered brand value** (a confirmed `.tokens.json` /
+   brand token) → **flag, never auto-apply** (`brand-source-of-truth`:
+   consumer brand wins). Precedence: brand tokens > confirmed `DESIGN.md` >
+   imported observation.
+5. On the human's accept, persist the chosen fields into `DESIGN.md`. Where the
+   consumer wants a token source of truth, hand the mapped DTCG fields to
+   [`brand-to-tokens`](../../brand-to-tokens/SKILL.md) / `design-tokens` to
+   materialise `.tokens.json` — do not invent a parallel token format.
+
+**Two sources, one shape:**
+
+- **External target** (a site/repo you don't own) → an external tool emits the
+  artifact; import it here.
+- **Current repo** → prefer [`existing-ui-audit`](../../existing-ui-audit/SKILL.md);
+  it already inventories the codebase and can emit the same `design-system.json`
+  shape, so the import path is identical either way.

package/dist/agent-src/skills/design-tokens/SKILL.md

@@ -13,6 +13,9 @@ trust:
   level: professional
 install:
   removable: true
+scope:
+  write: []
+  verification_reason: "execution declares no handler, so this skill runs nothing of its own — every write is the calling agent's, under the rules that govern it. No command can prove a scope the skill never executes."
 execution:
   type: manual
 ---

package/dist/agent-src/skills/design-variations/SKILL.md

@@ -12,6 +12,9 @@ trust:
   level: professional
 install:
   removable: true
+scope:
+  write: []
+  verification_reason: "execution declares no handler, so this skill runs nothing of its own — every write is the calling agent's, under the rules that govern it. No command can prove a scope the skill never executes."
 execution:
   type: manual
 ---

package/dist/agent-src/skills/developer-like-execution/SKILL.md

@@ -3,6 +3,9 @@ model_tier: inherit
 name: developer-like-execution
 description: "Use when implementing, debugging, refactoring, or reviewing code — enforces the think → analyze → verify → execute workflow — even when the user just says 'implement X' without naming it."
 domain: process
+scope:
+  write: []
+  verification_reason: "execution.handler is internal, so this skill spawns no subprocess — writes happen through the agent's declared allowed_tools. No command can prove a scope the skill never executes."
 execution:
   type: assisted
   handler: internal

package/dist/agent-src/skills/discovery-interview/SKILL.md

@@ -1,7 +1,7 @@
 ---
 model_tier: inherit
 name: discovery-interview
-description: "Use when running discovery interviews — question-bank build, bias audit, insight extraction. Triggers on 'audit my guide', 'extract insights from transcript', 'is my hypothesis falsifiable'."
+description: "When running discovery interviews — question-bank build, bias audit, insight extraction. Triggers on 'audit my guide', 'extract insights from transcript', 'is my hypothesis falsifiable'."
 status: active
 tier: senior
 domain: product

package/dist/agent-src/skills/doc-coauthoring/SKILL.md

@@ -1,7 +1,7 @@
 ---
 model_tier: inherit
 name: doc-coauthoring
-description: "Use when co-authoring a PRD, design doc, RFC, decision doc, or technical spec — 3-stage flow (context → section-by-section → reader-test) — even if the user just says 'help me write this spec'."
+description: "When co-authoring a PRD, design doc, RFC, decision doc, or technical spec — 3-stage flow (context → section-by-section → reader-test) — even if the user just says 'help me write this spec'."
 domain: process
 workspaces:
   - agent-config-maintainer

package/dist/agent-src/skills/docx-authoring/SKILL.md

@@ -1,7 +1,7 @@
 ---
 model_tier: medium
 name: docx-authoring
-description: "Use when generating or editing a Word .docx — create, fill a template, or edit body XML via a consumer library; round-trip validated. Triggers on 'generate a docx', 'fill this Word template'."
+description: "When generating or editing a Word .docx — create, fill a template, or edit body XML via a consumer library; round-trip validated. Triggers on 'generate a docx', 'fill this Word template'."
 status: active
 tier: senior
 domain: process

package/dist/agent-src/skills/editorial-calendar/SKILL.md

@@ -1,7 +1,7 @@
 ---
 model_tier: inherit
 name: editorial-calendar
-description: "Use when shaping cadence — evergreen / campaign / reactive split, beat-mapping across channel stages, content-debt management. Triggers on 'plan our content cadence', 'what should we publish'."
+description: "When shaping cadence — evergreen / campaign / reactive split, beat-mapping across channel stages, content-debt management. Triggers on 'plan our content cadence', 'what should we publish'."
 status: active
 tier: senior
 domain: product

package/dist/agent-src/skills/estimate-ticket/SKILL.md

@@ -10,6 +10,9 @@ personas:
   - critical-challenger
   - ai-agent
 domain: product
+scope:
+  write: []
+  verification_reason: "execution.handler is internal, so this skill spawns no subprocess — writes happen through the agent's declared allowed_tools. No command can prove a scope the skill never executes."
 execution:
   type: assisted
   handler: internal

package/dist/agent-src/skills/existing-ui-audit/SKILL.md

@@ -88,7 +88,7 @@ Detect markers, in order. **Stop at the first match** — projects rarely run mo
 | Marker | Signal | Where |
 |---|---|---|
 | Flux | `livewire/flux` in `composer.json`, `<flux:*>` tags in views | `composer.json`, grep `resources/views` |
-| shadcn/ui | `components.json` exists at repo root | `components.json` |
+| shadcn/ui | `components.json` exists at the **workspace** root — the package that owns it; in a monorepo this is never the repository root | `components.json` |
 | Headless UI | `@headlessui/react` or `@headlessui/vue` in `package.json` | `package.json` |
 | Radix | `@radix-ui/*` in `package.json` (without shadcn marker) | `package.json` |
 | Material/Chakra/Mantine/Ant | their package names in `package.json` | `package.json` |
@@ -125,7 +125,7 @@ read, so easing and duration stop being values the brief silently regenerates.
 
 ### 4. Detect shadcn inventory (only when `state.stack.frontend == "react-shadcn"`)
 
-Read `components.json` for the registered style + base color, then read `package.json` for `@radix-ui/*` and any locally vendored `components/ui/*.tsx` files. Write into `state.ui_audit.shadcn_inventory`:
+Read `components.json` for the registered style + base color, then read the `package.json` **of the scope root** — `state.stack.scope_root`, relative to the project root, empty when the project root *is* the scope — for `@radix-ui/*` or the unified `radix-ui` package, plus any locally vendored `components/ui/*.tsx` files. In a monorepo the repository root carries neither the marker nor the dependency, so reading it finds nothing. Write into `state.ui_audit.shadcn_inventory`:
 
 ```
 {
@@ -137,6 +137,23 @@ Read `components.json` for the registered style + base color, then read `package
 }
 ```
 
+### 4b. Prefer a live Storybook read over the hand-read inventory — when one is running
+
+When the project carries `@storybook/addon-mcp` **and** a Storybook is running, query it
+instead of reading files: `list-all-documentation` for the inventory, then `get-documentation`
+for the components that matter. **The live read wins; the hand-read inventory of step 4 is the
+fallback and is never removed** — an agent that cannot reach a running Storybook must still be
+able to inventory the library. The channel disappearing is normal, not an error.
+
+**React-only while the toolset is in preview.** Storybook's own MCP FAQ (docs **10.5**,
+`docs/ai/mcp/overview` § FAQ) states the documentation toolset supports React only during
+preview, so Vue, Angular, and Web Components take the file-read path. Stated here rather than
+discovered at runtime, because the failure otherwise looks like a broken MCP server.
+
+Discipline for what the live read returns is [`storybook-workshop`](../storybook-workshop/SKILL.md)'s
+— in particular: never use a prop the manifest does not document, and fetch the project's
+story instructions before writing a story.
+
 ### 5. List reusable patterns
 
 Categorize what already exists. Empty arrays are valid, never omit the keys.
@@ -264,7 +281,7 @@ tool produces, so the import path is uniform.
 
 - The model tends to skip the audit and start designing straight from the request — the dispatcher gate at `directives/ui/audit.ts` enforces "no design without audit findings". Never treat this skill as optional for non-trivial UI.
 - The model tends to misidentify a single Tailwind utility as a "design token" — tokens come from the config or `:root`, not from class strings in templates.
-- Don't assume a Radix-only `package.json` means shadcn — shadcn requires `components.json` at repo root.
+- Don't assume a Radix-only `package.json` means shadcn — shadcn requires `components.json` at the workspace root that owns the components. Each workspace carries its own; the repository root carries none, which is the layout the shadcn CLI scaffolds.
 - `state.ui_audit.shadcn_inventory.version` is often missing; the shadcn CLI does not always pin itself in `package.json`. Record `null` rather than guessing.
 - Greenfield is detected, not assumed — a project with one Blade layout and no components is still greenfield only if tokens AND design system markers AND components are all empty.
 - **There is no mtime cache.** An earlier revision told you to cache by

package/dist/agent-src/skills/fe-design/SKILL.md

@@ -22,6 +22,16 @@ dispatching this?**
 |---|---|---|
 | A ticketed run — the UI directive set is dispatching | `directives/ui/*` | a reference it cites |
 | Anything else — "build me the page", "improve this form", "implement the approved design" | **you, in this turn** | the executor |
+| A renderer axis is in play — WebGL / Three.js / canvas / scroll-scrubbed video | **still you** — the renderer is a grounding question, not a second executor | the executor, grounded via `search_stack` |
+
+The third row is an **axis**, not an owner. Renderer selection resolves through
+machinery that already exists: `search_stack` in
+[`corpus-grounding`](../corpus-grounding/SKILL.md)
+(`scripts/decision_engine.ts`), whose stack corpus carries `threejs.csv` among
+its stacks, read beside the register in
+[`design-intelligence`](../design-intelligence/references/context-and-registers.md)
+§ Register. No second frontend executor is declared, and none is needed: a
+renderer changes what you ground against, never who writes the UI.
 
 Outside the engine, nothing else owns the design quality of a UI write. Reading
 this skill and then writing the UI from priors is the failure it exists to
@@ -65,42 +75,36 @@ else runs the loop.
    artifact's own markup/CSS/JS is stack-compatible, adapt **that code**;
    re-deriving it from scratch is a deviation needing confirmation
    (`design-fidelity-mechanics` § Adopt the code).
-5. **Review** — run [`design-review`](../design-review/SKILL.md) before calling
-   it done, and scope the verdict honestly: render-scoped when you can render
-   it, otherwise explicitly static-scoped, naming which checks actually ran.
-   "Looks good" with neither scope named is a verdict without evidence.
-
-### The heuristics — here, not one hop away
-
-The load-bearing subset, inline so that "loaded" means the content is in
-context. Depth stays in [`references/design-patterns.md`](references/design-patterns.md).
-
-1. **Audit outranks heuristic.** A token, primitive or convention the audit
-   found wins over anything on this list; say `[audit override]` when it does.
-2. **Reuse tokens, never raw values.** A hex, font or px literal where the
-   audit found a token is off-brand by construction.
-3. **All five states are designed.** `empty` is a helpful message, not a blank
-   region; `loading` prefers a skeleton over a spinner (it shows structure, so
-   it reads as faster); `error` says what to do next.
-4. **Labels are always visible.** No placeholder-only inputs.
-5. **Validate on blur and on submit**, never on every keystroke; the message
-   goes below the field and is specific.
-6. **Mobile-first, and 320 px actually works.** Default styles are the small
-   viewport; complexity is added at larger breakpoints, never removed at
-   smaller ones.
-7. **A11y minimums are non-negotiable:** 4.5:1 contrast for text (3:1 large),
-   every interactive element reachable by Tab, a visible focus ring, semantic
-   elements over `div`, and an `aria-label` wherever there is no visible text.
-8. **Every action gets feedback.** Users read silence as failure.
-9. **Prefer undo over a confirmation dialog** — confirmations get clicked
-   through; an undo affordance actually protects the user.
-10. **Tables:** numbers right-aligned, text left-aligned, a designed empty
-    state, and a sticky header on anything long enough to scroll.
-11. **No placeholder microcopy ships.** Lorem ipsum, `TODO`, and
-    `[Your text here]` are unfinished output, not drafts.
-
-Before proposing a direction, run the anti-slop scan named under
-[Anti-slop discipline](#anti-slop-discipline) below.
+5. **Review, then re-enter** — run [`design-review`](../design-review/SKILL.md)
+   before calling it done, and scope the verdict honestly: render-scoped when
+   you can render it, otherwise explicitly static-scoped, naming which checks
+   actually ran. "Looks good" with neither scope named is a verdict without
+   evidence. Findings → fix them and **re-enter step 5**, at most **2 rounds**,
+   and **stop early on a null**: a round that produces **no new finding** ends
+   the loop, and a round that produces only findings the provided artifact
+   already covers produces no new finding. At the 2-round ceiling with findings
+   still open, stop and hand the remaining list back — ship-as-is or abort is
+   the user's call, never another silent pass. Judgement alone never buys a
+   third round.
+
+   The ceiling is **2**, the same number the engine enforces
+   (`directives/ui/polish.ts`, `POLISH_CEILING`), so the ad-hoc path and the
+   ticketed path bound the loop identically rather than by two conventions that
+   can drift. Fixtures: `daf-adhoc-converges` (round 2 is a null, loop ends) and
+   `daf-adhoc-ceiling` (findings remain at the ceiling, loop hands back).
+
+### The floors — pulled immediately before the write
+
+```
+PULL references/craft-floor.md IMMEDIATELY BEFORE THE WRITE, NOT AT SKILL LOAD.
+```
+
+Twelve universal floors that do not vary by mode, register, intent or stack.
+They were inline here until the delivery point moved: a floor read at skill-load
+time can be a compaction away from the write it governs (`ADR-227`).
+
+One precedence rule stays here because it is authority, not craft: **the audit
+outranks every heuristic.** Say `[audit override]` when it does.
 
 ## Positioning — reference inside the engine, executor outside it
 
@@ -212,13 +216,42 @@ When this skill's content is folded into a design brief or review:
 - Heuristics in this reference apply across stacks; do not promote them to project rules without checking the audit.
 - Mobile-first is not optional — every layout must work on 320px width.
 
+## Read the authority object — never re-infer it
+
+Read the resolved `ui_authority`
+([contract](../../../docs/contracts/ui-authority.md)) before the loop:
+`surface_mode`, `register`, `change_intent`, `reference_maturity`,
+`constraints`. This skill is a **declared consumer** — a second decision table
+beside the object is a drift surface, so do not re-derive any field.
+
+`surface_mode` sets density, hierarchy and expressiveness only, per
+[`design-modes`](../../../docs/guidelines/design-modes.md) § The second axis.
+
+```
+QUALITY FLOORS DO NOT VARY BY SURFACE MODE.
+A FLOOR THAT MOVES WITH THE MODE IS A PREFERENCE WEARING A FLOOR'S NAME.
+```
+
+### The source-led path
+
+`reference_maturity: runnable-artifact` → the artifact's own markup, CSS and JS
+is the data basis; adapting it is the **default** and a from-scratch
+re-derivation is a deviation. Sort every mechanic into `honoured` / `translated`
+/ `flagged`; one present, absent and unflagged is a **silent drop**, target zero.
+Procedure and the vanishing-mechanic classes:
+[`references/source-led-port.md`](references/source-led-port.md).
+
+Maturity is not decided here — it arrives on the authority object, and per-value
+provenance belongs to `road-to-frontend-fidelity-calibration` Phases 0 and 2.
+
 ## Anti-slop discipline
 
-Before proposing any UI layout, component, or aesthetic direction, pull
-[`docs/guidelines/design-antipatterns.md`](../../../docs/guidelines/design-antipatterns.md)
-and scan the Visual (V1–V7), Layout (L1–L8), and Quality-floors (Q1–Q12) sections.
-If the first-impulse design matches a listed pattern, either choose a different
-approach or explicitly invoke the override condition in the design brief.
+Before proposing a direction, scan the Visual (V1–V7) and Layout (L1–L8)
+sections of [`design-antipatterns`](../../../docs/guidelines/design-antipatterns.md).
+On a match, choose differently or invoke the entry's own override condition in
+the brief — every entry has one, and they are register-scoped. The Q* floors are
+in [`references/craft-floor.md`](references/craft-floor.md) instead of here, so
+they arrive at the write rather than at skill load.
 
 ## Do NOT
 

package/dist/agent-src/skills/fe-design/evals/triggers.json

@@ -1,55 +1,67 @@
 {
-    "skill": "fe-design",
-    "last_eval": "2026-08-12",
-    "description": "Should-trigger covers BOTH halves of this skill's two modes: planning/exploration (give me directions, explore, plan the flow, pattern choice) AND — new — ad-hoc implementation, because outside a ticketed run no other artefact owns the design quality of a UI write. The implementation query used to be labelled should-not-trigger on the reasoning that it belongs to the apply step; that step only exists as a dispatch target inside the engine, so the label routed ad-hoc implementation prompts to nobody. Should-not-trigger keeps the genuine near-misses: inventorying existing UI (existing-ui-audit owns it in both modes), a ui-trivial one-line edit, and non-UI work.",
-    "queries": [
-        {
-            "q": "give me three directions for the pricing page",
-            "trigger": true
-        },
-        {
-            "q": "explore a few visual directions for the dashboard before we commit",
-            "trigger": true
-        },
-        {
-            "q": "design the multi-step onboarding flow",
-            "trigger": true
-        },
-        {
-            "q": "should this filter be a modal or an inline panel?",
-            "trigger": true
-        },
-        {
-            "q": "plan the responsive layout for the settings screen",
-            "trigger": true
-        },
-        {
-            "q": "implement the selected checkout design in React exactly as approved",
-            "trigger": true
-        },
-        {
-            "q": "build the settings page",
-            "trigger": true
-        },
-        {
-            "q": "improve this form, it feels clunky",
-            "trigger": true
-        },
-        {
-            "q": "what form components do we already have in the codebase?",
-            "trigger": false
-        },
-        {
-            "q": "change the primary button colour to green",
-            "trigger": false
-        },
-        {
-            "q": "add a database index on orders.status",
-            "trigger": false
-        },
-        {
-            "q": "fix the failing unit test in the auth module",
-            "trigger": false
-        }
-    ]
+  "skill": "fe-design",
+  "last_eval": "2026-08-12",
+  "description": "Should-trigger covers BOTH halves of this skill's two modes: planning/exploration (give me directions, explore, plan the flow, pattern choice) AND \u2014 new \u2014 ad-hoc implementation, because outside a ticketed run no other artefact owns the design quality of a UI write. The implementation query used to be labelled should-not-trigger on the reasoning that it belongs to the apply step; that step only exists as a dispatch target inside the engine, so the label routed ad-hoc implementation prompts to nobody. Should-not-trigger keeps the genuine near-misses: inventorying existing UI (existing-ui-audit owns it in both modes), a ui-trivial one-line edit, and non-UI work. Scroll-storytelling rows added 2026-08-23: a scroll-driven narrative build is an ad-hoc UI write and routes here, and the renderer axis is a third row in this skill's mode table rather than a second executor. The should-NOT-trigger row is the one that matters \u2014 \"scroll position resets when I navigate back\" is a state/restoration bug and belongs to the frontend-engineer lens, not to design. Without it the word \"scroll\" alone would route every scroll bug to a design skill.",
+  "queries": [
+    {
+      "q": "give me three directions for the pricing page",
+      "trigger": true
+    },
+    {
+      "q": "explore a few visual directions for the dashboard before we commit",
+      "trigger": true
+    },
+    {
+      "q": "design the multi-step onboarding flow",
+      "trigger": true
+    },
+    {
+      "q": "should this filter be a modal or an inline panel?",
+      "trigger": true
+    },
+    {
+      "q": "plan the responsive layout for the settings screen",
+      "trigger": true
+    },
+    {
+      "q": "implement the selected checkout design in React exactly as approved",
+      "trigger": true
+    },
+    {
+      "q": "build the settings page",
+      "trigger": true
+    },
+    {
+      "q": "improve this form, it feels clunky",
+      "trigger": true
+    },
+    {
+      "q": "what form components do we already have in the codebase?",
+      "trigger": false
+    },
+    {
+      "q": "change the primary button colour to green",
+      "trigger": false
+    },
+    {
+      "q": "add a database index on orders.status",
+      "trigger": false
+    },
+    {
+      "q": "fix the failing unit test in the auth module",
+      "trigger": false
+    },
+    {
+      "q": "build the scroll-driven story section for the launch page",
+      "trigger": true
+    },
+    {
+      "q": "make the hero pin and scrub through three beats as you scroll",
+      "trigger": true
+    },
+    {
+      "q": "the scroll position resets when I navigate back \u2014 fix the restoration bug",
+      "trigger": false
+    }
+  ]
 }

package/dist/agent-src/skills/fe-design/references/craft-floor.md

@@ -0,0 +1,52 @@
+# Craft floor — load this immediately before the write
+
+Universal floors. They do not vary by surface mode, register, change intent or
+stack. Taste and direction stay mode-scoped in
+[`fe-design`](../SKILL.md); nothing on this page is a preference.
+
+**Why this is a separate page rather than inline.** The skill body used to carry
+these inline, on the argument that "loaded" should mean the content is in
+context. That argument is kept and the delivery point moved: a floor read at
+*skill-load* time can be many turns and a compaction away from the write it
+governs, and `ADR-227` records that a path-scoped carrier is not re-injected
+after `/compact`. This page is pulled **immediately before the write**, which is
+the moment the floor has to be present. Same content, later delivery.
+
+## The floors
+
+1. **Reuse tokens, never raw values.** A hex, font or px literal where the audit
+   found a token is off-brand by construction.
+2. **All five states are designed.** `empty` is a helpful message, not a blank
+   region. `loading` prefers a skeleton over a spinner — it shows structure, so
+   it reads as faster. `error` says what to do next.
+3. **Labels are always visible.** No placeholder-only inputs.
+4. **Validate on blur and on submit**, never on every keystroke. The message
+   goes below the field and is specific.
+5. **Mobile-first, and 320 px actually works.** Default styles are the small
+   viewport; complexity is added at larger breakpoints, never removed at smaller
+   ones. `agent-config ui:render` captures 320 px so this is measurable rather
+   than asserted.
+6. **Accessibility minimums are non-negotiable.** 4.5:1 contrast for text (3:1
+   large), every interactive element reachable by Tab, a visible focus ring,
+   semantic elements over `div`, and an `aria-label` wherever there is no
+   visible text.
+7. **Every action gets feedback.** Users read silence as failure.
+8. **Prefer undo over a confirmation dialog.** Confirmations get clicked
+   through; an undo affordance actually protects the user.
+9. **Tables:** numbers right-aligned, text left-aligned, a designed empty state,
+   and a sticky header on anything long enough to scroll.
+10. **No placeholder microcopy ships.** Lorem ipsum, `TODO` and
+    `[Your text here]` are unfinished output, not drafts.
+11. **A default font pick is declared or changed.** Inter, Roboto, DM Sans,
+    Geist, Space Grotesk and Instrument Serif are the AI defaults. In the
+    product register "one reliable family" is a complete reason — but state it.
+    An undeclared pick is an undeclared choice (T7).
+12. **Zero silent drops.** A mechanic present in a supplied source and absent
+    from the output, with no `flagged` entry, is a silent drop.
+
+## What is NOT on this page
+
+Density, hierarchy, expressiveness, palette character, type personality,
+motion register. Those are surface-mode and register decisions and they live in
+`fe-design`. Putting them here would make a preference look like a floor, which
+is the failure this split exists to prevent.