docs: align lookback test purpose and spec normalization list

Honest test docstring (old/new semantics equivalent on contiguous
entries; test locks post-fix behavior against future regressions),
and add severity-prefix strip to the spec's normalization list.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

docs/superpowers/specs/2026-05-04-pz-deterministic-classifier-design.md

@@ -126,6 +126,7 @@ signature   = sha256(pattern_id + mod_id)[:16]
 
 Normalization for `pattern_id`:
 - Strip session metadata prefix (`General  f:N, t:N, st:N,N,N,N>` shape)
+- Strip body-prefix severity token (`ERROR:` / `SEVERE:` / `WARN:` / `FATAL:`, case-insensitive) so a body that opens with the severity word still hashes the same as one that doesn't.
 - Flatten double- and single-quoted strings to `"<S>"` / `'<S>'`
 - Flatten ≥2-digit numeric runs to `<N>`
 - Collapse whitespace
@@ -240,5 +241,6 @@ Test invocation: `python -m unittest discover tools/pz-analyzer/tests/` should b
 - Editing `pz_error_analysis.py` or `pz_redact_all.sh`.
 - Modifying any file in `/opt/ik-codex/src/`, `/opt/ik-codex/test/`, or `/opt/iblogs/`.
 - AI / LLM integration of any kind in the new tool.
+- LLM inference at runtime in iblogs / bosslogs production. The Qwen analyzer (`pz_error_analysis.py`) is a developer-only discovery tool used to expand the deterministic ruleset in `pz_parser.py` (and its future PHP port). Production rendering is deterministic-only, forever.
 - iblogs front-end rendering of the classification output.
 - Filesystem mod-scan reattribution (pzmm's symbol/vehicle indexes).

tools/pz-analyzer/tests/test_attribution.py

@@ -169,12 +169,15 @@ class RawLineLookbackTests(unittest.TestCase):
             path.unlink()
 
     def test_multiline_entry_does_not_shrink_practical_lookback(self) -> None:
+        """Multi-line entries inside the lookback window do not break
+        attribution. (Old body-line-budget and new raw-line-distance semantics
+        happen to be equivalent on contiguous PZ entries; this test locks the
+        post-fix semantic against future regression to a budget that *would*
+        differ — e.g. a body-line cap with a smaller value.)
+        """
         # Layout the file so a multi-line entry sits between marker and ERROR.
-        # Under the OLD body-line-budget semantics the multi-line entry's 5
-        # continuation lines would consume the budget and push the marker
-        # outside the window. Under raw-line semantics the marker on line 1 is
-        # still within 40 raw lines of the ERROR even though the file has a
-        # 6-line multi-line entry in between.
+        # The marker on line 1 is within 40 raw lines of the ERROR even though
+        # the file has a 6-line multi-line entry in between.
         lines = [_make_marker_line(0)]            # raw line 1: marker entry
         # Single-line fillers on raw lines 2..30 (29 entries).
         for i in range(1, 30):
@@ -207,9 +210,11 @@ class RawLineLookbackTests(unittest.TestCase):
             self.assertEqual(entries[-1].line_start, 41)
             records = pz_parser.classify_entries(entries, source_file="ml.txt")
             self.assertEqual(len(records), 1)
-            # Under the OLD body-line-budget rule, the 5 stack-frame lines
-            # plus the surrounding fillers would have pushed the marker out
-            # of the budget. Under raw-line semantics it survives.
+            # Raw-line-distance semantics: the marker on line 1 is 40 raw
+            # lines from the ERROR on line 41, so attribution holds. (Old
+            # body-line-budget would also pass here on contiguous entries;
+            # this assertion locks the post-fix behavior against future
+            # regression to a tighter cap.)
             self.assertEqual(records[0].attribution, "inferred")
             self.assertEqual(records[0].mod_id, "testmodalpha")
         finally: