fix: review rubric knows all 15 libraries

.github/workflows/impl-review.yml

@@ -327,6 +327,8 @@ jobs:
             - SPEC_ID: ${{ steps.pr.outputs.specification_id }}
             - PR_NUMBER: ${{ steps.pr.outputs.pr_number }}
             - ATTEMPT: ${{ steps.attempts.outputs.display }}
+            - LANGUAGE: ${{ steps.lang.outputs.language }}
+            - EXT: ${{ steps.lang.outputs.ext }}
 
       - name: Extract quality score
         id: score

CHANGELOG.md

@@ -28,6 +28,19 @@ aggregate instead: an italic *Catalog* line at the end of the version section an
   MuiDivider, MuiSkeleton, and MuiAlert are now wired to the CSS-var system
   (`--ink-soft`/`--rule`/`--bg-elevated`), and the two `borderColor: 'divider'` usages in
   SpecTabs/RelatedSpecs use `var(--rule)` (audit 2026-07-08 High#4) (#9622).
+- **Review rubric finally knows all 15 libraries** — the workflow-active reviewer prompt and
+  scoring criteria still described the 9-Python-library era: the file-extension guide called
+  highcharts Python and knew no `.js`/`.tsx`, SC-04's title rule rejected `javascript` (and in
+  `quality-criteria.md` even `julia`) as a language token — silently costing every correct
+  JS/Julia title points — `${LANGUAGE}`/`${EXT}` were referenced in the prompt but never passed
+  by `impl-review.yml`, the interactive-fairness rules (don't penalize hover/zoom invisible in
+  the PNG) lived only in a file no workflow reads, and the step-10 `review_checklist.json`
+  example still showed the pre-rebalance 5-category shape (`library_features`, 40/25/20/10/5
+  maxima) that corrupted ~6% of stored checklists. All lists are corrected to the canonical
+  registry, the fairness rules (plus a JS-harness note for CQ-05) moved into the running prompt,
+  AR-05 gained rows for chartjs/d3/echarts/muix, and new registry-derived tests in
+  `tests/unit/prompts/` fail on any future library-list drift (audit 2026-07-08 Medium#1 +
+  Medium#12) (#9623).
 
 ### Changed
 

prompts/quality-criteria.md

@@ -37,7 +37,7 @@ Checks that gate quality scoring. On fail: Score=0. Workflow-handled checks (AR-
 
 | ID | Check | Description | Verification |
 |----|-------|-------------|--------------|
-| AR-01 | SYNTAX_ERROR | Code cannot be parsed | `python -m py_compile` |
+| AR-01 | SYNTAX_ERROR | Code cannot be parsed | Language parser (`python -m py_compile`, `Rscript -e parse`, `node --check`, …) |
 | AR-02 | RUNTIME_ERROR | Code throws exception | Execution with timeout |
 | AR-03 | NO_OUTPUT | No `plot.png` created | File exists? |
 | AR-04 | EMPTY_PLOT | Image empty | < 10KB or > 95% white |
@@ -61,8 +61,12 @@ Implementation must use **plot functions** from the library, not just styling.
 | altair | `alt.Chart().mark_*()` | Only `configure_*()` |
 | plotnine | `ggplot() + geom_*()` | Only `theme()` |
 | pygal | Chart classes with data | Only config |
-| highcharts | Chart with series | Only options |
+| highcharts | `Highcharts.chart()` with series | Only options objects |
 | letsplot | `ggplot() + geom_*()` | Only `ggsize()` |
+| chartjs | `new Chart(...)` with datasets | Only options/plugin config |
+| d3 | Data joins + scales rendering marks (`selection.data(...).join(...)`) | Only selections/styling |
+| echarts | `chart.setOption(...)` with series | Only `echarts.init()`/theme config |
+| muix | MUI X chart components (`<LineChart>`, `<ScatterChart>`, …) with series | Only `ThemeProvider`/styling |
 
 **Note:** Using data loading utilities (e.g., `sns.load_dataset()`, `sklearn.datasets`) from other libraries is allowed and does not count as library usage. This check only evaluates whether the implementation uses the library's **plotting functions**.
 
@@ -72,7 +76,7 @@ When a library cannot technically implement a spec (e.g., pygal cannot do 3D), t
 
 ### AR-08: Fake Functionality
 
-A static library (matplotlib, seaborn, plotnine) simulates interactive features that cannot work in a PNG image.
+A static library (matplotlib, seaborn, plotnine, ggplot2, makie) simulates interactive features that cannot work in a PNG image.
 
 **Triggers (auto-reject):**
 - Simulated tooltips (annotation boxes styled to look like hover tooltips)
@@ -358,7 +362,7 @@ This category evaluates aesthetic sophistication beyond mere correctness. A plot
 
 | Points | Criterion |
 |--------|-----------|
-| 3 | Title is `{spec-id} · {language} · {library} · anyplot.ai`, optionally prefixed with `{Descriptive Title} · ` (language ∈ {python, r}). Legend labels correct |
+| 3 | Title is `{spec-id} · {language} · {library} · anyplot.ai`, optionally prefixed with `{Descriptive Title} · ` (language ∈ {python, r, julia, javascript}). Legend labels correct |
 | 2 | Title format correct but legend issues, or vice versa |
 | 1 | Partially correct |
 | 0 | Missing or wrong |

prompts/quality-evaluator.md

@@ -4,7 +4,7 @@
 
 ## Role
 
-You are a strict code reviewer for data visualizations. Most implementations are Python; ggplot2 is R; Makie.jl is Julia. You evaluate plot implementations against `prompts/quality-criteria.md`.
+You are a strict code reviewer for data visualizations. Most implementations are Python; ggplot2 is R; Makie.jl is Julia; chartjs, d3, echarts, and highcharts are JavaScript; muix is JavaScript (React, TSX). You evaluate plot implementations against `prompts/quality-criteria.md`.
 
 ## Two-Stage Evaluation
 
@@ -44,7 +44,7 @@ You evaluate implementations that passed all auto-reject checks. Focus purely on
 ## Input
 
 1. **Specification**: From `plots/{spec-id}/specification.md`
-2. **Code**: From `plots/{spec-id}/implementations/{language}/{library}{ext}` — `{ext}` is `.py` for python libraries, `.R` for ggplot2, and `.jl` for makie
+2. **Code**: From `plots/{spec-id}/implementations/{language}/{library}{ext}` — `{ext}` is `.py` for Python libraries, `.R` for ggplot2, `.jl` for makie, `.js` for the JavaScript libraries (chartjs, d3, echarts, highcharts), and `.tsx` for muix
 3. **Previews**: BOTH theme renders of the plot image — `plot-light.png` and `plot-dark.png`. You must inspect both. For interactive libraries, also `plot-light.html` and `plot-dark.html`.
 4. **Library Rules**: From `prompts/library/{library}.md`
 5. **Style Guide** (canonical palette + theme tokens): `prompts/default-style-guide.md` — consult its "Categorical Palette", "Continuous Data", "Background", and "Theme-adaptive Chrome" sections for VQ-07 scoring.
@@ -234,7 +234,7 @@ Not AR-09 (handle via VQ-05 instead): text overflowing its axis but staying on t
 | SC-01 | Plot Type | 5 | Correct chart type? |
 | SC-02 | Required Features | 4 | All spec features present? |
 | SC-03 | Data Mapping | 3 | X/Y correctly assigned? All data visible? |
-| SC-04 | Title & Legend | 3 | Title is `{spec-id} · {language} · {library} · anyplot.ai`, optionally prefixed with `{Descriptive Title} · ` (language ∈ {python, r, julia}). Legend labels correct? |
+| SC-04 | Title & Legend | 3 | Title is `{spec-id} · {language} · {library} · anyplot.ai`, optionally prefixed with `{Descriptive Title} · ` (language ∈ {python, r, julia, javascript}). Legend labels correct? |
 
 ### Step 4: Data Quality (15 pts)
 
@@ -316,7 +316,7 @@ Both have ~5.76M pixels, so font size recommendations apply equally.
 
 **Important**: Interactive libraries also produce HTML output alongside PNG.
 
-**Interactive libraries**: plotly, bokeh, altair, highcharts, pygal, letsplot
+**Interactive libraries**: altair, bokeh, chartjs, d3, echarts, highcharts, letsplot, muix, plotly, pygal
 
 **DO NOT penalize** interactive features that aren't visible in PNG:
 - HoverTool, tooltips, hover info

prompts/workflow-prompts/ai-quality-review.md

@@ -17,7 +17,7 @@ Evaluate if the **${LIBRARY}** implementation matches the specification for `${S
 - Note all required features
 
 ### 2. Read the Implementation
-`plots/${SPEC_ID}/implementations/${LANGUAGE}/${LIBRARY}${EXT}` — `${EXT}` is `.py` for python libraries (matplotlib, seaborn, plotly, bokeh, altair, plotnine, pygal, highcharts, letsplot), `.R` for ggplot2, and `.jl` for makie
+`plots/${SPEC_ID}/implementations/${LANGUAGE}/${LIBRARY}${EXT}` — the workflow supplies `${LANGUAGE}` and `${EXT}` for this run. Reference: `.py` for the Python libraries (matplotlib, seaborn, plotly, bokeh, altair, plotnine, pygal, letsplot), `.R` for ggplot2 (R), `.jl` for makie (Julia), `.js` for the JavaScript libraries (chartjs, d3, echarts, highcharts), and `.tsx` for muix (JavaScript, React)
 
 ### 3. Read Library-Specific Rules
 `prompts/library/${LIBRARY}.md`
@@ -104,6 +104,15 @@ Visually estimate from each PNG — no pixel measurement needed. These are soft
 - "Title spans ~80% of width at fontsize=14pt." → Expected for the long mandated anyplot title; no deduction.
 - "Y-axis label 'Fläche von Häusern in Quadratmetern' takes ~40% of axis length at fontsize=12pt." → Genuinely long label at sensible fontsize; no deduction as long as it doesn't overflow the axis.
 
+### 5e. Interactive & JavaScript Libraries — Fairness Rules
+
+**Interactive libraries** (produce `plot-{theme}.html` alongside the PNG): altair, bokeh, chartjs, d3, echarts, highcharts, letsplot, muix, plotly, pygal.
+
+- **DO NOT penalize** interactive features that aren't visible in the PNG — tooltips/hover info, zoom/pan/selection tools, interactive legends, crossfiltering. The PNG is a static preview; these features add real value in the HTML detail view. Treat well-configured interactivity as a strength (e.g. "Uses HoverTool for detailed data inspection in HTML output").
+- **Only criticize** interactivity when it hurts the static render: hover-only labels that leave the PNG unreadable, misconfigured features, or errors.
+
+**JavaScript libraries** (chartjs, d3, echarts, highcharts as `.js`; muix as `.tsx`): the snippet renders into the browser harness's pre-sized `#container` mount node and never saves files itself — the harness (`automation/js-render/render.mjs`) captures `plot-{theme}.png` and `plot-{theme}.html`. For CQ-05, check the mount-node contract and current library API instead of `savefig`-style calls; animations must be disabled (`animation: false` or the library's equivalent) so the harness doesn't screenshot mid-animation.
+
 ### 6. Check for Auto-Reject (AR-08, AR-09)
 
 **AR-08 — Fake interactivity (static libraries only — matplotlib, seaborn, plotnine, ggplot2, makie):**
@@ -167,7 +176,7 @@ Read `prompts/quality-criteria.md` and evaluate:
 | SC-01 | Plot Type | 5 | Correct chart type? |
 | SC-02 | Required Features | 4 | All features from spec? |
 | SC-03 | Data Mapping | 3 | X/Y correct? Axes show all data? |
-| SC-04 | Title & Legend | 3 | Title is `{spec-id} · {language} · {library} · anyplot.ai`, optionally prefixed with `{Descriptive Title} · ` (language ∈ {python, r, julia}). Legend labels match? |
+| SC-04 | Title & Legend | 3 | Title is `{spec-id} · {language} · {library} · anyplot.ai`, optionally prefixed with `{Descriptive Title} · ` (language ∈ {python, r, julia, javascript}). Legend labels match? |
 
 #### Data Quality (15 pts)
 | ID | Criterion | Max | Check |
@@ -324,21 +333,26 @@ Dark render (plot-dark.png):
   Legibility verdict: PASS | FAIL (explain if FAIL)
 EOF
 
-# Criteria checklist as structured JSON
+# Criteria checklist as structured JSON.
+# Use EXACTLY these six keys — visual_quality, design_excellence,
+# spec_compliance, data_quality, code_quality, library_mastery — with these
+# maxima (30/20/15/15/10/10, matching step 7). The website's quality tab and
+# the stored metadata rely on this exact shape; no other keys, no renames.
 cat > review_checklist.json << 'EOF'
 {
   "visual_quality": {
-    "score": 36,
-    "max": 40,
+    "score": 24,
+    "max": 30,
     "items": [
-      {"id": "VQ-01", "name": "Text Legibility", "score": 10, "max": 10, "passed": true, "comment": "All text readable"},
-      {"id": "VQ-02", "name": "No Overlap", "score": 8, "max": 8, "passed": true, "comment": "No overlapping elements"}
+      {"id": "VQ-01", "name": "Text Legibility", "score": 7, "max": 8, "passed": true, "comment": "All text readable in both themes"},
+      {"id": "VQ-02", "name": "No Overlap", "score": 5, "max": 6, "passed": true, "comment": "No collisions"}
     ]
   },
-  "spec_compliance": {"score": 23, "max": 25, "items": [...]},
-  "data_quality": {"score": 18, "max": 20, "items": [...]},
-  "code_quality": {"score": 10, "max": 10, "items": [...]},
-  "library_features": {"score": 5, "max": 5, "items": [...]}
+  "design_excellence": {"score": 12, "max": 20, "items": [...]},
+  "spec_compliance": {"score": 13, "max": 15, "items": [...]},
+  "data_quality": {"score": 12, "max": 15, "items": [...]},
+  "code_quality": {"score": 9, "max": 10, "items": [...]},
+  "library_mastery": {"score": 6, "max": 10, "items": [...]}
 }
 EOF
 ```

tests/unit/prompts/test_prompts.py

@@ -14,6 +14,8 @@
 
 import pytest
 
+from core.constants import INTERACTIVE_LIBRARIES, LANGUAGE_FILE_EXTENSIONS, LIBRARIES_METADATA, SUPPORTED_LANGUAGES
+
 
 # Base paths
 PROMPTS_DIR = Path(__file__).parent.parent.parent.parent / "prompts"
@@ -419,3 +421,77 @@ def test_code_blocks_have_language(self) -> None:
                     f"{filepath.name} has too few labeled code blocks "
                     f"({labeled}/{total} = {labeled_ratio:.0%}, need ≥20%)"
                 )
+
+
+class TestFifteenLibraryCoverage:
+    """Review prompts must know the full library registry.
+
+    Guards against the drift the 2026-07-08 audit found (M#12): scoring
+    prompts still describing the 9-Python-library era while the pipeline
+    reviews R, Julia, and JavaScript implementations. Expectations derive
+    from core.constants so a future library lands here as a failing test,
+    not as silent scoring bias.
+    """
+
+    WORKFLOW_PROMPTS_DIR = PROMPTS_DIR / "workflow-prompts"
+
+    SCORING_PROMPTS = [PROMPTS_DIR / "quality-evaluator.md", PROMPTS_DIR / "workflow-prompts" / "ai-quality-review.md"]
+
+    @pytest.mark.parametrize("prompt_path", SCORING_PROMPTS, ids=lambda p: p.name)
+    def test_interactive_library_list_matches_registry(self, prompt_path: Path) -> None:
+        """The '**Interactive libraries**' line must EQUAL the canonical set —
+        a missing entry biases scoring against that library, an extra entry
+        makes reviewers expect HTML output the library never produces."""
+        content = prompt_path.read_text()
+        line = next((line for line in content.splitlines() if line.startswith("**Interactive libraries**")), None)
+        assert line is not None, f"{prompt_path.name} has no '**Interactive libraries**' line"
+        listed = {entry.strip().rstrip(".") for entry in line.rsplit(":", 1)[1].split(",")}
+        assert listed == set(INTERACTIVE_LIBRARIES), (
+            f"{prompt_path.name} interactive list drifted — "
+            f"missing: {sorted(set(INTERACTIVE_LIBRARIES) - listed)}, "
+            f"extra: {sorted(listed - set(INTERACTIVE_LIBRARIES))}"
+        )
+
+    @pytest.mark.parametrize("prompt_path", SCORING_PROMPTS, ids=lambda p: p.name)
+    def test_every_file_extension_documented(self, prompt_path: Path) -> None:
+        """Language defaults plus per-library overrides (.tsx) must all appear."""
+        expected = set(LANGUAGE_FILE_EXTENSIONS.values()) | {
+            lib["file_extension"] for lib in LIBRARIES_METADATA if "file_extension" in lib
+        }
+        content = prompt_path.read_text()
+        missing = [ext for ext in sorted(expected) if f"`{ext}`" not in content]
+        assert not missing, f"{prompt_path.name} missing file extensions: {missing}"
+
+    @pytest.mark.parametrize(
+        "prompt_path",
+        [
+            PROMPTS_DIR / "quality-criteria.md",
+            PROMPTS_DIR / "quality-evaluator.md",
+            PROMPTS_DIR / "workflow-prompts" / "ai-quality-review.md",
+        ],
+        ids=lambda p: p.name,
+    )
+    def test_sc04_title_rule_accepts_every_language(self, prompt_path: Path) -> None:
+        """SC-04's language set must include every supported language, or every
+        correct JS/Julia title costs points."""
+        content = prompt_path.read_text()
+        sc04_lines = [line for line in content.splitlines() if "language ∈" in line]
+        assert sc04_lines, f"{prompt_path.name} has no SC-04 'language ∈' rule"
+        joined = "\n".join(sc04_lines)
+        missing = [lang for lang in sorted(SUPPORTED_LANGUAGES) if lang not in joined]
+        assert not missing, f"{prompt_path.name} SC-04 language set missing: {missing}"
+
+    def test_checklist_example_uses_canonical_six_keys(self) -> None:
+        """The step-10 review_checklist.json example must show exactly the six
+        canonical category keys (the website renders this shape directly)."""
+        content = (self.WORKFLOW_PROMPTS_DIR / "ai-quality-review.md").read_text()
+        for key in (
+            "visual_quality",
+            "design_excellence",
+            "spec_compliance",
+            "data_quality",
+            "code_quality",
+            "library_mastery",
+        ):
+            assert f'"{key}"' in content, f"checklist example missing key: {key}"
+        assert "library_features" not in content, "stale 5-category checklist key resurfaced"