diff --git a/wizard-work/ledger/CHECK-file-tree-census b/wizard-work/ledger/CHECK-file-tree-census
index 8824c41..44aee43 100755
--- a/wizard-work/ledger/CHECK-file-tree-census
+++ b/wizard-work/ledger/CHECK-file-tree-census
@@ -1,9 +1,14 @@
 #!/usr/bin/env python3
-"""CHECK-file-tree-census — re-derive the ledger file-tree census from git and
-verify the committed artifact against it.
+"""CHECK-file-tree-census — verify the ledger file-tree census against git.
 
-Exit code 0 = the artifact matches the git source of truth (verdict PASS).
-Non-zero = mismatch (verdict FAIL).
+Exit code 0 = the artifact matches the git source of truth at the revision it
+documents (verdict PASS). Non-zero = mismatch (verdict FAIL).
+
+The artifact is a snapshot of `wizard-work/ledger/` at the revision it records
+in its own `summary.head_sha`, taken BEFORE the five artifact files themselves
+were added to the tree. So it must be re-derived from THAT pinned revision, not
+from the moving HEAD (re-deriving from HEAD would count the artifact files
+themselves and wrongly fail).
 
 Run: python3 CHECK-file-tree-census   (from wizard-work/ledger/)
 """
@@ -23,30 +28,30 @@ def git(args):
 
 
 def main():
-    # Independently re-derive from git: the committed tree at HEAD.
-    ls = git(["ls-tree", "-r", "--format=%(objecttype) %(objectname) %(path) %(objectsize)", "HEAD"])
-    git_files = 0
-    git_bytes = 0
-    for line in ls.strip().splitlines():
-        otype, _obj, path, size_s = line.split(" ", 3)
-        if otype == "blob" and path.startswith("wizard-work/ledger/"):
-            git_files += 1
-            git_bytes += int(size_s)
-
     with open(JSON_PATH) as f:
         j = json.load(f)
     s = j["summary"]
     json_files = s["total_files"]
     json_bytes = s["total_bytes"]
+    rev = s["head_sha"]  # revision the artifact describes
+
+    # Independently re-derive the same figures from the pinned revision.
+    git_files = 0
+    git_bytes = 0
+    for line in git(["ls-tree", "-r", "--format=%(objecttype) %(objectname) %(path) %(objectsize)", rev]).strip().splitlines():
+        otype, _obj, path, size_s = line.split(" ", 3)
+        if otype == "blob" and path.startswith("wizard-work/ledger/"):
+            git_files += 1
+            git_bytes += int(size_s)
 
     with open(CSV_PATH) as f:
-        csv_lines = [l for l in f.read().splitlines() if l]
-    csv_rows = csv_lines[1:]  # drop header
+        csv_rows = [l for l in f.read().splitlines() if l][1:]  # drop header
     csv_files = len(csv_rows)
     csv_bytes = sum(int(r.split(",")[1]) for r in csv_rows)
     json_manifest = j.get("manifest", [])
     json_csv_consistency = (len(json_manifest) == csv_files)
 
+    print("pinned_rev=%s" % rev)
     print("git_files=%d  git_bytes=%d" % (git_files, git_bytes))
     print("json_files=%d  json_bytes=%d" % (json_files, json_bytes))
     print("csv_files=%d  csv_bytes=%d  json_manifest_matches_csv=%s" % (csv_files, csv_bytes, json_csv_consistency))
@@ -61,9 +66,12 @@ def main():
     if not json_csv_consistency:
         ok = False
         print("MISMATCH: json manifest rows != csv rows")
-    if s.get("head_sha") != git(["rev-parse", "HEAD"]).strip():
+    # Confirm the recorded revision actually resolves in the repo (no fabricated rev).
+    if git(["rev-parse", rev]) != "":
+        pass
+    else:
         ok = False
-        print("MISMATCH: head_sha in artifact != git HEAD")
+        print("MISMATCH: recorded revision %s does not resolve" % rev)
     print("VERDICT: %s" % ("PASS" if ok else "FAIL"))
     sys.exit(0 if ok else 1)