fix: scan harness/, and make the orphan check real rather than claimed

Both findings from the review on #33, and the second is the sharper one.

The checker skipped harness/opfs-contract/src/lib.rs — a shipped Rust source
tree with 14 runs over the limit, including a 22-line module doc — so it could
have reported success while the convention was violated. harness/ is now in the
default roots.

And design-principles claimed a renamed test was made detectable because 'the
checker looks for both halves'. It did not: it counted comment lines and listed
doctest fences, and nothing read context/test-evidence/ at all. So the cost the
convention says is controlled was uncontrolled, in the same commit that
introduced the convention — a reviewed paragraph is not evidence, and this one
was mine.

--check-evidence now requires every '### name' heading to name a test that
exists. One direction only: a test may have no evidence (most do not), but
evidence naming a test nobody can find is a dangling explanation.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

scripts/check-comment-runs.py

@@ -51,14 +51,64 @@ def rust_files(roots):
                     yield os.path.join(dirpath, f)
 
 
+EVIDENCE_DIR = os.path.join("context", "test-evidence")
+
+
+def check_evidence(roots):
+    """Every `### name` in context/test-evidence/ must name a test that exists.
+
+    One direction only, deliberately: a test may have no evidence (most do
+    not), but evidence naming a test nobody can find is an orphan -- which is
+    the cost design-principles names for splitting the narrative from the test.
+    """
+    if not os.path.isdir(EVIDENCE_DIR):
+        print("no context/test-evidence/ yet: nothing to check")
+        return 0
+
+    claimed = {}
+    for dirpath, _, filenames in os.walk(EVIDENCE_DIR):
+        for f in sorted(filenames):
+            if not f.endswith(".md"):
+                continue
+            path = os.path.join(dirpath, f)
+            for n, raw in enumerate(open(path, encoding="utf8", errors="replace"), 1):
+                s = raw.strip()
+                if s.startswith("### "):
+                    claimed.setdefault(s[4:].strip().strip("`"), []).append(f"{path}:{n}")
+
+    defined = set()
+    for path in rust_files(roots):
+        for raw in open(path, encoding="utf8", errors="replace"):
+            s = raw.strip()
+            if s.startswith("fn ") or s.startswith("pub fn ") or s.startswith("async fn "):
+                name = s.split("fn ", 1)[1].split("(")[0].split("<")[0].strip()
+                defined.add(name)
+
+    orphans = {k: v for k, v in claimed.items() if k not in defined}
+    if not orphans:
+        print(f"test-evidence: {len(claimed)} entries, all name a test that exists")
+        return 0
+    print(f"{len(orphans)} test-evidence entries name no test:\n", file=sys.stderr)
+    for name, where in sorted(orphans.items()):
+        print(f"  {name}  ({', '.join(where)})", file=sys.stderr)
+    return 1
+
+
 def main():
     ap = argparse.ArgumentParser()
-    ap.add_argument("roots", nargs="*", default=["src", "crates", "tests", "examples"])
+    ap.add_argument("roots", nargs="*",
+                    default=["src", "crates", "tests", "examples", "harness"])
     ap.add_argument("--limit", type=int, default=LIMIT)
     ap.add_argument("--list-doctests", action="store_true",
                     help="list doc comments containing a ``` fence and exit")
+    ap.add_argument("--check-evidence", action="store_true",
+                    help="every context/test-evidence heading must name a test that exists")
     args = ap.parse_args()
-    roots = [r for r in (args.roots or ["src", "crates", "tests", "examples"]) if os.path.isdir(r)]
+    default_roots = ["src", "crates", "tests", "examples", "harness"]
+    roots = [r for r in (args.roots or default_roots) if os.path.isdir(r)]
+
+    if args.check_evidence:
+        return check_evidence(roots)
 
     if args.list_doctests:
         for path in sorted(rust_files(roots)):