docs: update docs and templates for category reclassification

.claude/agents/calibration/converter.md

@@ -113,7 +113,7 @@ Write results to `$RUN_DIR/conversion.json`.
   "uncoveredStruggles": [
     {
       "description": "A difficulty not covered by any flagged rule",
-      "suggestedCategory": "layout | token | component | naming | ai-readability | handoff-risk",
+      "suggestedCategory": "structure | token | component | naming | behavior",
       "estimatedImpact": "easy | moderate | hard | failed"
     }
   ]

.claude/agents/rule-discovery/designer.md

@@ -21,7 +21,7 @@ You will receive:
    - Read `src/core/rules/rule-config.ts` for score/severity conventions
 3. Design the rule:
    - **Rule ID**: kebab-case, descriptive (e.g., `missing-component-description`)
-   - **Category**: existing (`layout | token | component | naming | ai-readability | handoff-risk`) or propose a new category if none fits. New categories require justification.
+   - **Category**: existing (`structure | token | component | naming | behavior`) or propose a new category if none fits. New categories require justification.
    - **Severity**: `blocking | risk | missing-info | suggestion`
    - **Initial score**: based on estimated impact on implementation difficulty
    - **Check logic**: what condition triggers the violation

.coderabbit.yaml

@@ -94,8 +94,8 @@ reviews:
         Ensure workflows use pnpm/action-setup@v4 and --frozen-lockfile.
 
   pre_merge_checks:
-    docstring_coverage:
-      enabled: false
+    docstrings:
+      mode: "off"
 
 chat:
   auto_reply: true

.github/ISSUE_TEMPLATE/bug.yml

@@ -7,7 +7,7 @@ body:
     attributes:
       label: Symptom
       description: What's happening? Include error messages, wrong output, or unexpected behavior.
-      placeholder: "invisible-layer scores -10 (blocking) but hidden layers don't block implementation"
+      placeholder: "unnecessary-node scores -10 (blocking) but hidden layers don't block implementation"
     validations:
       required: true
   - type: textarea
@@ -31,7 +31,7 @@ body:
     attributes:
       label: Affected Files
       description: Which source files need changes?
-      placeholder: "src/core/rules/rule-config.ts, src/core/rules/ai-readability/index.ts"
+      placeholder: "src/core/rules/rule-config.ts, src/core/rules/structure/index.ts"
     validations:
       required: false
   - type: textarea

docs/CALIBRATION.md

@@ -171,11 +171,11 @@ Real excerpts from calibration debates across 5 fixtures and 7+ rounds.
 > Converter (with filtered nodes): "GROUP inside a real UI component causes moderate difficulty, not hard. The hard ratings were from icon nodes."
 > Critic: "High confidence, 3 cases. 37.5% change within limit. Severity back to risk."
 
-### ambiguous-structure: validated at -10
+### no-auto-layout: validated with absorbed rules
 
-**Consistently confirmed across all fixtures:**
+**After absorbing `ambiguous-structure` and `missing-layout-hint`:**
 > Converter: "Children named 'Path', 'Path', 'Path' with no semantic distinction — programmatic conversion is impossible without visual rendering."
-> Critic: "4 cases, all hard. Score -10 at blocking confirmed. No change needed."
+> Critic: "4 cases, all hard. The merged `no-auto-layout` rule now covers these structure clarity signals. Score -7 at blocking confirmed."
 
 ### New rule proposals: VECTOR no path data
 

docs/MCP-VS-CLI.md

@@ -23,12 +23,11 @@ This report documents a side-by-side comparison of analyzing the same Figma desi
 | Category | MCP | CLI | Match |
 |----------|-----|-----|-------|
 | **Overall** | C (67%) | C (67%) | Yes |
-| Layout | 100% | 100% | Yes |
+| Structure | 100% | 100% | Yes |
 | Token | 21% | 21% | Yes |
 | Component | 100% | 100% | Yes |
 | Naming | 24% | 24% | Yes |
-| AI Readability | 59% | 59% | Yes |
-| Handoff Risk | 100% | 100% | Yes |
+| Behavior | 100% | 100% | Yes |
 
 ### Issue Counts
 
@@ -37,7 +36,7 @@ This report documents a side-by-side comparison of analyzing the same Figma desi
 | **Total Issues** | **5** | **6** | **No** |
 | Token issues | 2 | 3 | No |
 | Naming issues | 2 | 2 | Yes |
-| AI Readability issues | 1 | 1 | Yes |
+| Structure issues | 1 | 1 | Yes |
 
 ### Issues by Rule
 
@@ -46,7 +45,7 @@ This report documents a side-by-side comparison of analyzing the same Figma desi
 | `raw-color` | **1** | **2** | **No** |
 | `inconsistent-spacing` | 1 | 1 | Yes |
 | `default-name` | 2 | 2 | Yes |
-| `empty-frame` | 1 | 1 | Yes |
+| `unnecessary-node` | 1 | 1 | Yes |
 
 ---
 

docs/REFERENCE.md

@@ -92,18 +92,19 @@ Override score, severity, or enable/disable individual rules:
 
 ### All Rule IDs
 
-**Layout (9 rules)**
+**Structure (9 rules)**
 
 | Rule ID | Default Score | Default Severity |
 |---------|--------------|-----------------|
 | `no-auto-layout` | -7 | blocking |
 | `absolute-position-in-auto-layout` | -10 | blocking |
-| `fixed-width-in-responsive-context` | -4 | risk |
 | `missing-responsive-behavior` | -4 | risk |
 | `group-usage` | -5 | risk |
 | `fixed-size-in-auto-layout` | -5 | risk |
-| `missing-min-width` | -5 | risk |
-| `missing-max-width` | -4 | risk |
+| `missing-size-constraint` | -5 | risk |
+| `unnecessary-node` | -2 | suggestion |
+| `z-index-dependent-layout` | -5 | risk |
+| `deep-nesting` | -4 | risk |
 
 **Token (7 rules)**
 
@@ -117,13 +118,14 @@ Override score, severity, or enable/disable individual rules:
 | `raw-opacity` | -5 | risk |
 | `multiple-fill-colors` | -3 | missing-info |
 
-**Component (3 rules)**
+**Component (4 rules)**
 
 | Rule ID | Default Score | Default Severity |
 |---------|--------------|-----------------|
 | `missing-component` | -7 | risk |
 | `detached-instance` | -5 | risk |
 | `missing-component-description` | -2 | missing-info |
+| `variant-structure-mismatch` | -4 | risk |
 
 **Naming (5 rules)**
 
@@ -135,23 +137,14 @@ Override score, severity, or enable/disable individual rules:
 | `numeric-suffix-name` | -2 | missing-info |
 | `too-long-name` | -1 | suggestion |
 
-**AI Readability (5 rules)**
+**Behavior (4 rules)**
 
 | Rule ID | Default Score | Default Severity |
 |---------|--------------|-----------------|
-| `ambiguous-structure` | -10 | blocking |
-| `z-index-dependent-layout` | -5 | risk |
-| `missing-layout-hint` | -3 | missing-info |
-| `invisible-layer` | -1 | suggestion |
-| `empty-frame` | -2 | missing-info |
-
-**Handoff Risk (3 rules)**
-
-| Rule ID | Default Score | Default Severity |
-|---------|--------------|-----------------|
-| `hardcode-risk` | -5 | risk |
 | `text-truncation-unhandled` | -5 | risk |
-| `prototype-link-in-design` | -2 | suggestion |
+| `prototype-link-in-design` | -2 | missing-info |
+| `overflow-behavior-unknown` | -3 | missing-info |
+| `wrap-behavior-unknown` | -3 | missing-info |
 
 ### Example Configs
 
@@ -178,12 +171,13 @@ Override score, severity, or enable/disable individual rules:
 }
 ```
 
-**Mobile-first (focus on layout):**
+**Mobile-first (focus on structure):**
+
 ```json
 {
   "rules": {
     "missing-responsive-behavior": { "score": -10, "severity": "blocking" },
-    "fixed-width-in-responsive-context": { "score": -8, "severity": "blocking" },
+    "fixed-size-in-auto-layout": { "score": -8, "severity": "blocking" },
     "no-auto-layout": { "score": -12 }
   }
 }
@@ -207,7 +201,7 @@ Each custom rule is a JSON object with the following fields:
 | Field | Type | Required | Description |
 |-------|------|----------|-------------|
 | `id` | `string` | Yes | Unique rule identifier (kebab-case recommended) |
-| `category` | `string` | Yes | One of: `layout`, `token`, `component`, `naming`, `ai-readability`, `handoff-risk` |
+| `category` | `string` | Yes | One of: `structure`, `token`, `component`, `naming`, `behavior` |
 | `severity` | `string` | Yes | One of: `blocking`, `risk`, `missing-info`, `suggestion` |
 | `score` | `number` | Yes | Penalty score (integer, must be <= 0) |
 | `match` | `object` | Yes | Conditions to evaluate (see below) |
@@ -317,7 +311,7 @@ Frames with many children that should have Auto Layout applied:
 ```json
 {
   "id": "large-frame-no-auto-layout",
-  "category": "layout",
+  "category": "structure",
   "severity": "risk",
   "score": -5,
   "match": {
@@ -341,7 +335,7 @@ Hidden nodes buried deep in the tree that add noise:
 ```json
 {
   "id": "deep-hidden-layer",
-  "category": "ai-readability",
+  "category": "structure",
   "severity": "suggestion",
   "score": -2,
   "match": {
@@ -418,7 +412,7 @@ I want a custom rule for CanICode that checks: [DESCRIBE WHAT TO CHECK]
 
 Generate a JSON rule object with these fields:
 - id: kebab-case identifier
-- category: one of layout, token, component, naming, ai-readability, handoff-risk
+- category: one of structure, token, component, naming, behavior
 - severity: one of blocking, risk, missing-info, suggestion
 - score: negative integer (more negative = more severe)
 - match: object with conditions (ALL must match). Available conditions:

docs/SCORING.md

@@ -1,7 +1,7 @@
 # Scoring Model
 
-canicode scores designs on a 0-100% scale across 6 categories:
-**layout**, **token**, **component**, **naming**, **ai-readability**, **handoff-risk**.
+canicode scores designs on a 0-100% scale across 5 categories:
+**structure**, **token**, **component**, **naming**, **behavior**.
 
 ## How Scores Are Calculated
 
@@ -31,7 +31,7 @@ Issues concentrated in one rule type are easier to fix than scattered issues acr
 
 ```
 category_score = density * 0.7 + diversity * 0.3
-overall_score  = average of all 6 category scores (equal weight)
+overall_score  = average of all 5 category scores (equal weight)
 ```
 
 The 70:30 ratio prioritizes volume over variety. A design with a single systemic problem (e.g., all frames missing auto-layout) is still very hard to implement even though diversity is low.

docs/fixtures/ANALYSIS-VALIDITY-FEEDBACK.md

@@ -21,12 +21,12 @@ It is **not** a statistical proof. It assumes: REST fixture JSON + current `rule
 
 ### A. Invisible layers + raw color clusters (many Simple DS fixtures)
 
-- **Overrated:** Overall **grade can look decent (A / high B)** while the tree still carries many **invisible-layer** and **raw-color** hits. Severity may keep the headline grade up, but the AI still sees **noise** (extra nodes, ambiguous fills) and may **mis-order or mis-style** visible content. **Pixel outcome** can be worse than the letter grade suggests.
+- **Overrated:** Overall **grade can look decent (A / high B)** while the tree still carries many **unnecessary-node** and **raw-color** hits. Severity may keep the headline grade up, but the AI still sees **noise** (extra nodes, ambiguous fills) and may **mis-order or mis-style** visible content. **Pixel outcome** can be worse than the letter grade suggests.
 - **Underrated:** Rare for the **same** pattern; occasionally invisible only affects editor hygiene, not export — but REST + rules still complain.
 - **Remedies:**
   - **Figma:** Delete or detach truly unused layers; replace raw values with variables where the team cares; reduce hidden decoration.
   - **Workflow:** Feed **`design-tree`** or a **pruned node list** to the AI so it ignores known-noise IDs.
-  - **Tooling:** Consider reporting **“visible-only issues”** aggregate for AI-facing summaries; tune `invisible-layer` / `raw-color` weights after calibration for “impact on visual-compare.”
+  - **Tooling:** Consider reporting **“visible-only issues”** aggregate for AI-facing summaries; tune `unnecessary-node` / `raw-color` weights after calibration for “impact on visual-compare.”
 
 ### B. Unscoped large trees (Material 3 & large Simple DS)
 
@@ -48,7 +48,7 @@ It is **not** a statistical proof. It assumes: REST fixture JSON + current `rule
 
 ### D. Numeric / generic naming (Material 3 community)
 
-- **Overrated:** Score can look **OK on layout** while **ai-readability / naming** tanks — sometimes **underweighted** in how much it confuses **less capable** models (variable names, component names in code).
+- **Overrated:** Score can look **OK on layout** while **structure / naming** tanks — sometimes **underweighted** in how much it confuses **less capable** models (variable names, component names in code).
 - **Underrated:** For **vision-heavy** codegen (screenshot + structure), **bad names** matter less — score may **undersell** “actually buildable if you ignore names.”
 - **Remedies:**
   - **Figma:** Rename critical interactive nodes; keep DS naming convention.
@@ -83,7 +83,7 @@ Names refer to directories under `fixtures/` or `fixtures/done/`. Grades refer t
 
 ### `simple-ds-175-9106` / `175-8591` / `175-7790` / `562-9518` — B ~ B+
 
-- **Overrated:** **raw-color + invisible-layer** volume — **B/B+** may **overrate** first-shot pixel match.
+- **Overrated:** **raw-color + unnecessary-node** volume — **B/B+** may **overrate** first-shot pixel match.
 - **Underrated:** **Detach/instance** and **default-name** sometimes fixable with a strong “use Figma structure as source of truth” prompt.
 - **Remedies:** Scope; variable cleanup; explicit **instance** handling in prompt.