Merge integration/rpa-dashboard-skills into ticket 04

Keeps both sides: frontend detection and bot wiring from tickets 02/03, optional
sections from ticket 04. The a-empty-sections eval now uses the shared "fixture" key.

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
@@ -38,7 +39,7 @@ FIXTURES = {
         "sections": [],
     },
     "a-empty-sections": {
-        "dir": "a-empty",
+        "fixture": "a-empty",
         "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. Colours: Superbill Ember. "
                   "Desktop or web: web. Optional sections: all three (checkpoint tracker, admin/roles "
@@ -48,14 +49,87 @@ FIXTURES = {
         "never_asks": [r"\bkeep\b|restyle"],
         "sections": ["checkpoints", "admin", "login"],
     },
+    "b-pyqt-bot": {
+        "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. Colours: Tebra Sky. "
+                  "Optional sections: none.",
+        "frontend": r"^\s*(none|no\b)",
+        "asks": [r"colou?r", r"section"],
+        "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):
+    """Bot fixtures: the page is wired, and a dry run of the generated wiring gives real steps and logs.
+    Returns the env that points the page check at that run."""
+    checks["page wired to the bot (sample data off)"] = bool(re.search(r"SAMPLE_DATA\s*=\s*false", html))
+    runners = [p for p in project.rglob("bot_runner.py") if ".claude" not in p.parts]
+    checks["bot_runner.py built"] = len(runners) == 1
+    if not runners:
+        return {}
+    host = [runners[0], runners[0].with_name("bridge.py")]
+    checks["no placeholders left in the host"] = all(p.exists() and "{{" not in p.read_text(encoding="utf-8") for p in host)
+    dry = subprocess.run([sys.executable, str(runners[0])], cwd=project, capture_output=True, text=True,
+                         encoding="utf-8", timeout=300)
+    replay.write_text(dry.stdout, encoding="utf-8")
+    events = [json.loads(x) for x in dry.stdout.splitlines() if x.startswith("{")]
+    steps = [e["data"] for e in events if e["event"] == "step"]
+    checks["bot dry run ends done"] = bool(events) and events[-1]["data"].get("state") == "done"
+    checks["real steps reach the page"] = len({s["label"] for s in steps}) >= 3 and steps[-1]["pct"] == 100
+    checks["bot error line reaches the log"] = any(e["event"] == "log" and e["data"]["type"] == "error" for e in events)
+    return {"PAGE_CHECK_WIRED": "1", "PAGE_CHECK_REPLAY": str(replay)}
 
 
 def main(fixture):
     cfg = FIXTURES[fixture]
     work = Path(tempfile.mkdtemp(prefix=f"rpa-eval-{fixture}-"))
     project = work / "project"
-    shutil.copytree(ROOT / "tests" / "fixtures" / cfg.get("dir", 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)
 
@@ -73,12 +147,21 @@ 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)
@@ -99,8 +182,13 @@ def main(fixture):
             for name, trace in traces.TRACES.items():
                 want = name in cfg["sections"]
                 checks[f"{name} {'built' if want else 'left no trace'}"] = bool(trace.search(html)) == want
+        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"))
         pc = subprocess.run([sys.executable, "-m", "pytest", str(ROOT / "tests" / "page_check"), "-q"],
-                            env={**os.environ, "PAGE_CHECK_HTML": str(pages[0])}, capture_output=True, text=True)
+                            env=env, capture_output=True, text=True)
         checks["page check passes"] = pc.returncode == 0
         page_check = pc.stdout.strip().splitlines()[-1:] if pc.stdout else [pc.stderr[-500:]]
         if cfg.get("sections"):  # every chosen section's page check ran