diff --git a/docs/superpowers/specs/2026-07-09-mouse-postprocess-quality-design.md b/docs/superpowers/specs/2026-07-09-mouse-postprocess-quality-design.md
index e1b122b..dda5903 100644
--- a/docs/superpowers/specs/2026-07-09-mouse-postprocess-quality-design.md
+++ b/docs/superpowers/specs/2026-07-09-mouse-postprocess-quality-design.md
@@ -89,8 +89,13 @@ post-processing pipeline alone meets the quality gates (max tail turn
    change; scroll golden untouched). CHANGELOG entry.
 2. **Unit tests** (`tests/unit/test_postprocess.py`) — backtrack
    tolerance, overshoot compression, exact endpoint hit after global
-   correction, correction weights 0/1 at the ends, no turns >90° on
-   smoothed output.
+   correction, correction weights 0/1 at the ends. The tail-quality
+   guard allows at most ONE >90° reversal (the natural
+   overshoot-and-correct gesture that overshoot support implies); two
+   or more indicate the hook/zigzag artifact class. (Amended
+   2026-07-09: the original "no turns >90°" wording predated overshoot
+   support and was empirically over-strict — a single 92-135° reversal
+   appears in ~20% of traces and is correct behavior.)
 3. **Acceptance** — re-run the diagnostic script (same 4 cases × 6
    seeds) and compare: `turns>45°` count drops sharply, no vertical
    wall in the last 10 points. Final gate: user visually approves the