Merge integration/rpa-dashboard-skills (ticket 03) into ticket 02

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/evals/run_eval.py

@@ -9,6 +9,7 @@ prompt (`-p` can't answer questions). It then grades the reply, runs the Seam 1
 check on the page it built, and writes tests/evals/results/<fixture>.json.
 Exit code 0 means every check passed.
 """
+import difflib
 import importlib.util
 import json
 import os
@@ -45,7 +46,49 @@ FIXTURES = {
         "never_asks": [r"\bkeep\b|restyle", r"desktop|web"],
         "bot": True,
     },
+    # (c) existing teal frontend in web/, run once per answer to keep-or-restyle.
+    "c-keep": {
+        "fixture": "c-web-existing-frontend",
+        "prompt": "Build me an RPA dashboard for this project. I can't answer questions while you "
+                  "work, so here are my answers to anything the skill asks. Keep or restyle the "
+                  "existing frontend: keep its look. Optional sections: none.",
+        "frontend": r"(?=.*\bweb\b)(?=.*html)(?=.*(#0f766e|--color-brand))",
+        "asks": [r"keep|restyle", r"section"],
+        "never_asks": [r"^(?!.*(keep|restyle)).*(colou?r|Tebra Sky|Ember)", r"desktop"],
+        "page_has": ["--accent:#0f766e", "Trebuchet MS"],
+        "existing": "unchanged",
+    },
+    "c-restyle": {
+        "fixture": "c-web-existing-frontend",
+        "prompt": "Build me an RPA dashboard for this project. I can't answer questions while you "
+                  "work, so here are my answers to anything the skill asks. Keep or restyle the "
+                  "existing frontend: restyle it. Colours: Superbill Ember. Optional sections: none.",
+        "frontend": r"(?=.*\bweb\b)(?=.*html)(?=.*(#0f766e|--color-brand))",
+        "asks": [r"keep|restyle", r"Tebra Sky|Ember", r"section"],
+        "never_asks": [r"desktop"],
+        "page_has": ["--accent:#ff3049"],
+        "existing": "#ff3049",  # only token lines may change, and the theme now uses this colour
+    },
 }
+# A line a restyle may change: a custom property, a comment, a blank, or a web-font link.
+TOKEN_LINE = r"^\s*(--[\w-]+\s*:.*|/\*.*\*/|<link[^>]*fonts\.googleapis\.com.*|@import.*fonts\.googleapis\.com.*)?\s*$"
+
+
+def squash(s):
+    """Ignore spacing and case when looking for a token in the built page."""
+    return re.sub(r"\s+", "", s).lower()
+
+
+def changed_lines(src, project):
+    """Lines removed or added in the fixture's own files."""
+    out = []
+    for f in src.rglob("*"):
+        if f.is_file():
+            new = project / f.relative_to(src)
+            after = new.read_text(encoding="utf-8").splitlines() if new.exists() else []
+            out += [x[1:] for x in difflib.unified_diff(f.read_text(encoding="utf-8").splitlines(), after, lineterm="", n=0)
+                    if x[:1] in "+-" and x[:3] not in ("+++", "---")]
+    return out
 
 
 def bot_checks(project, html, checks, replay):
@@ -73,7 +116,8 @@ def main(fixture):
     cfg = FIXTURES[fixture]
     work = Path(tempfile.mkdtemp(prefix=f"rpa-eval-{fixture}-"))
     project = work / "project"
-    shutil.copytree(ROOT / "tests" / "fixtures" / fixture, project)
+    src = ROOT / "tests" / "fixtures" / cfg.get("fixture", fixture)
+    shutil.copytree(src, project)
     for skill in ROOT.glob("rpa-dashboard-*"):
         shutil.copytree(skill, project / ".claude" / "skills" / skill.name)
 
@@ -91,12 +135,21 @@ def main(fixture):
         return m[1] if m else ""
 
     checks = {"frontend reported correctly": bool(re.search(cfg["frontend"], line("FRONTEND"), re.I))}
+    questions = line("QUESTIONS").split("|")
     for pat in cfg["asks"]:
-        checks[f"asks /{pat}/"] = bool(re.search(pat, line("QUESTIONS"), re.I))
+        checks[f"asks /{pat}/"] = any(re.search(pat, q, re.I) for q in questions)
     for pat in cfg["never_asks"]:
-        checks[f"never asks /{pat}/"] = not re.search(pat, line("QUESTIONS"), re.I)
+        checks[f"never asks /{pat}/"] = not any(re.search(pat, q, re.I) for q in questions)
+    if "existing" in cfg:
+        changed = changed_lines(src, project)
+        if cfg["existing"] == "unchanged":
+            checks["existing files unchanged"] = not changed
+        else:
+            checks["existing files: only token lines changed"] = bool(changed) and all(re.match(TOKEN_LINE, x) for x in changed)
+            checks[f"existing tokens now use {cfg['existing']}"] = any(cfg["existing"] in x.lower() for x in changed)
 
-    pages = [p for p in project.rglob("*.html") if ".claude" not in p.parts]
+    pages = [p for p in project.rglob("*.html")
+             if ".claude" not in p.parts and not (src / p.relative_to(project)).exists()]  # new pages only
     checks["exactly one page built"] = len(pages) == 1
     results = ROOT / "tests" / "evals" / "results"
     results.mkdir(exist_ok=True)
@@ -110,6 +163,8 @@ def main(fixture):
             spec.loader.exec_module(palette)
             t = palette.derive(cfg["brand"])
             checks["custom palette applied"] = all(f"--accent:{t[m]['--accent']}" in html for m in t)
+        for needle in cfg.get("page_has", []):
+            checks[f"page has {needle}"] = squash(needle) in squash(html)
         env = {**os.environ, "PAGE_CHECK_HTML": str(pages[0])}
         if cfg.get("bot"):
             env.update(bot_checks(project, html, checks, results / f"{fixture}.run.jsonl"))