docs(ax): separate the paths axis from the base axis with a controlled middle point

@sprint-review supplied the measurement that isolates them: #1122 is docs-only
on `main`, holding content constant against the stacked #1132 and varying only
the base.

Measured live 2026-08-25:

  #1170  backend/  base main      11
  #1122  docs/     base main      10
  #1132  docs/     base stacked    4

#1170 → #1122 moves exactly one check, `E2E Tests` — the paths axis.
#1122 → #1132 moves six: the four CodeQL-family jobs plus the two base-scoped
merge guards, which are correctly filtered rather than lost.

So what stacking costs is the CodeQL family, entire and nothing else. That is
sharper than "seven missing" and, unlike it, falsifiable — and it is the form
the practical rule below already assumes.

Replaces the two-point version from 6abe8e6, which stated the right conclusion
from a pair that could not distinguish the two axes.

Docs-only.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: lilyshen0722

docs/development/agent-experience-audit.md

@@ -2535,17 +2535,35 @@ scanned. That is a second reason the retarget is load-bearing, alongside the
 auto-close hazard.
 
 **Not every absence is a stacking absence, and E2E is the one that fools you.**
-Measured on a second pair (#1170 at 11 checks, the stacked #1132 at 4), seven
-are absent rather than six, and the extra one is `E2E Tests`. It is tempting to
-add it to the stacking gap; it does not belong there. `playwright.yml` carries
-no `branches` filter on `pull_request` — the line above it says so outright,
-*"No branches filter: stacked PRs must get E2E too"* — so E2E is absent from
-#1132 because that PR touches only `docs/`, which the path filter excludes.
-@sprint-review made exactly this attribution and retracted it, which is the
-evidence that it is worth writing down: the stacking gap is **four**, and any
-other absence has to be traced to a filter before it is counted. Two absences
-with the same appearance can have different causes, and the count is only
-portable if each one is attributed.
+Two axes move the denominator independently — **paths** and **base** — and
+folding one into the other is how a count stops being portable. Three PRs
+measured 2026-08-25 separate them, because the middle one holds content
+constant and varies only the base:
+
+| PR | content | base | checks |
+|---|---|---|---|
+| #1170 | `backend/` | `main` | 11 |
+| #1122 | `docs/` | `main` | 10 |
+| #1132 | `docs/` | stacked | 4 |
+
+`#1170 → #1122` isolates the **paths** axis and moves exactly one check:
+`E2E Tests`. `playwright.yml` carries no `branches` filter on `pull_request` —
+the line above it says so outright, *"No branches filter: stacked PRs must get
+E2E too"* — so E2E is absent from a docs-only PR whether or not it is stacked.
+
+`#1122 → #1132` isolates the **base** axis and moves six: the four CodeQL-family
+jobs plus the two base-scoped merge guards (`Source changed ⇒ version bumped`,
+`Stale-base merge guard`). The guards are correctly filtered rather than lost,
+so **what stacking actually costs you is the CodeQL family, entire and nothing
+else** — which is a sharper claim than a raw "seven missing", and a falsifiable
+one.
+
+@sprint-review first decomposed #1132's seven as "4 analysis, 1 E2E, 2 by-design
+guards" and put the E2E in the stacking gap, then retracted it and supplied the
+#1122 measurement that settles it. Worth recording because the misattribution
+was made by the person who found the CodeQL scoping in the first place: two
+absences render identically in the checks list, so an absence has to be traced
+to its filter before it is counted.
 
 **The scoping of those three cannot be checked from inside a checkout.** The
 entry above notes there is no `codeql.yml`; the sharper form is that the