diff --git a/docs/DOC_TRUST_MAP.md b/docs/DOC_TRUST_MAP.md
index 0770d4f..28c03bc 100644
--- a/docs/DOC_TRUST_MAP.md
+++ b/docs/DOC_TRUST_MAP.md
@@ -108,6 +108,19 @@ the first sixteen lines of the file, and one carrying `Status` without
 `Review trigger` is reported as incomplete — that combination looks finished and
 is not.
 
+**Governs** is a comma-separated list, and an entry may explain itself after the
+glob with a spaced dash — `docs/data/** — the assets privacyllc.dev renders for
+this project`. `scripts/doc-triggers.py` cuts the entry there and reads the globs
+from the left of it. **Use that form and no other.** A gloss in parentheses, or
+after a colon, is not recognised: the whole entry becomes the glob, matches
+nothing, and the document is silently never fired — not reported as skipped
+either, because it still looks like a path. Three documents here sat in exactly
+that state from the first commit, `docs/data/img/README.md` among them, so
+editing the branding assets never once prompted the document that specifies their
+names and sizes. A trailing `/` on a glob means the directory and everything in
+it, which is how this file's `.githooks/` entry reaches a project's installed
+hooks.
+
 **Fires on** is optional, and only for the case where `Governs:` is much broader
 than the trigger. `Governs:` says *where* a document is authoritative; the
 trigger says *which kinds of change* to that place should bring somebody back,
diff --git a/docs/architecture/scripts/doc-triggers.py b/docs/architecture/scripts/doc-triggers.py
index 459fe9b..fa9e70a 100755
--- a/docs/architecture/scripts/doc-triggers.py
+++ b/docs/architecture/scripts/doc-triggers.py
@@ -155,6 +155,36 @@ def looks_like_path(glob: str) -> bool:
     return "/" in glob or "*" in glob or re.search(r"\.\w{1,5}$", glob) is not None
 
 
+# A `Governs:` entry is split on commas, so an entry that explains itself after
+# the glob arrives whole: `docs/data/** — the assets privacyllc.dev renders for
+# this project`. Used as a glob that matches nothing, ever, and because it
+# contains a slash `looks_like_path` calls it a path — so the document was
+# neither fired nor listed among the ones no change can fire mechanically. It was
+# simply absent, which is the one outcome a reader cannot notice.
+#
+# Three of the seven path-governing documents here were in that state from the
+# first commit, `docs/data/img/README.md` among them: editing the branding assets
+# had never once prompted the document that specifies their names and sizes.
+GLOSS = re.compile(r"\s+(?:—|–|--)\s+")
+
+
+def globs_in(entry: str) -> list[str]:
+    """The globs inside one `Governs:` entry, with any trailing gloss removed.
+
+    The cut requires whitespace on both sides of the dash: `source-grep` and
+    `doc-claims` appear in these headers and a bare `-` would halve them. Tokens
+    are taken from the left of the gloss rather than from the whole entry,
+    because prose on the right can itself look like a path — `privacyllc.dev`
+    passes the extension test and would become a glob that fires on a file
+    nobody has.
+
+    An entry yielding no token falls back to itself, so a shape not foreseen here
+    behaves exactly as it did before.
+    """
+    head = GLOSS.split(entry, 1)[0]
+    return [tok for tok in head.split() if looks_like_path(tok)] or [entry]
+
+
 def matches(path: str, glob: str) -> bool:
     """Whether `path` is governed by `glob`.
 
@@ -167,6 +197,10 @@ def matches(path: str, glob: str) -> bool:
         return False
     if glob.endswith("/**"):
         return path.startswith(glob[:-2]) or path == glob[:-3]
+    if glob.endswith("/"):
+        # A bare directory, as `githooks/README.md` governs ".githooks/". fnmatch
+        # would not match a file inside it.
+        return path.startswith(glob)
     if "/**/" in glob:
         head, tail = glob.split("/**/", 1)
         return path.startswith(head + "/") and fnmatch.fnmatch(path, "*" + tail)
@@ -320,8 +354,11 @@ def main() -> int:
         if not governs:
             continue
 
-        globs = [g.strip() for g in governs.split(",") if g.strip()]
-        path_globs = [g for g in globs if looks_like_path(g)]
+        entries = [g.strip() for g in governs.split(",") if g.strip()]
+        # Classification reads the whole entry and extraction reads inside it:
+        # deciding "path or subject?" on a token would move documents between the
+        # two lists as a side effect of this fix.
+        path_globs = [g for e in entries if looks_like_path(e) for g in globs_in(e)]
         if not path_globs:
             subject_only.append(rel)
             continue