quality: scan the framework, not just tools/, in the enumeration guard (#1017)

The guard's population is 'basename matches *test*.sh'. That is not a tools/
property, but Direction A scanned only framework/tools/, so a suite in a SIBLING
of tools/ was structurally invisible to the guard whose stated purpose is making
that impossible.

The live specimen, found by @scooby on origin/next:
framework/systemd/user/test-fleet-units.sh — a population member by the guard's
own definition, named on neither package.json nor ci.yml, and in no exclusion.
It is the only such file today. Same shape as #1017's original 17-of-39,
reintroduced for one subtree.

Three symmetric changes. The scan root moves to the framework. The S2 ci.yml
regex widens with it — not cosmetic: S1 is already general, S2 was the only
tools-scoped surface, so without it 'enumerate the file' would not be a
reachable disposition and an exclusion would be the only way to green. And the
directory-exists precondition follows the root it guards.

Disposition for the specimen: enumerated, not excluded. Its tmux block
self-skips on `command -v tmux && cc`; the rest — the unit-file assertions and
`systemd-analyze verify --user` — is real structural coverage that runs in CI
today. Measured in a CI shape (/usr/bin minus tmux, cc, systemd-analyze): rc=0.
That is strictly better than the two send-message siblings at exclusion lines
31-32, which have no CI-valuable tmux-free half.

Needles n9/c5 added, since a widening with no needle is the same silence one
layer up. n9 fails against the original guard. c5's scope is narrower than it
looks and the comment records the measurement: it passes vacuously on the
original and discriminates against the half-patch (scan wide, S2 narrow), which
is the realistic future regression.

    guard, real tree:  OK — population 53 (was 52), enumerated 38, excluded 15
    needles:           16 passed, 0 failed

packages/mosaic/framework/tools/quality/scripts/check-test-enumeration.sh

@@ -57,12 +57,18 @@ done
 PKG_JSON="$ROOT/packages/mosaic/package.json"
 CI_YML="$ROOT/.woodpecker/ci.yml"
 TOOLS_DIR="$ROOT/packages/mosaic/framework/tools"
+# The population is "basename matches *test*.sh", which is not a tools/ property.
+# Direction A used to scan TOOLS_DIR, so a suite in a SIBLING of tools/ was invisible
+# to the guard whose whole purpose is making that impossible — measured on origin/next
+# as systemd/user/test-fleet-units.sh, on no CI surface and in no exclusion (@scooby).
+# Scan the framework, so the scanned surface matches the claimed property.
+FRAMEWORK_DIR="$ROOT/packages/mosaic/framework"
 EXCLUSIONS="$TOOLS_DIR/quality/test-enumeration-exclusions.txt"
 
 for f in "$PKG_JSON" "$CI_YML"; do
     [[ -f "$f" ]] || { echo "FAIL: required surface file missing: $f" >&2; exit 2; }
 done
-[[ -d "$TOOLS_DIR" ]] || { echo "FAIL: tools dir missing: $TOOLS_DIR" >&2; exit 2; }
+[[ -d "$FRAMEWORK_DIR" ]] || { echo "FAIL: framework dir missing: $FRAMEWORK_DIR" >&2; exit 2; }
 
 fail_count=0
 fail() { printf 'FAIL  %s\n' "$1"; fail_count=$(( fail_count + 1 )); }
@@ -90,7 +96,7 @@ print("\n".join(seen))
 PY
 )
 
-# --- Surface 2: ci.yml, every framework/tools token wherever it appears ------
+# --- Surface 2: ci.yml, every framework token wherever it appears ------------
 # Comment lines (first non-whitespace char is #) are skipped BEFORE matching:
 # commenting an invocation out is the most common way a suite actually gets
 # disabled, and a raw-text regex would keep calling it enumerated (F1, 20155 on
@@ -98,7 +104,7 @@ PY
 # in a TRAILING comment on a live line still matches; no such line exists today
 # and full fidelity would need a YAML parser the CI image does not ship.
 mapfile -t S2 < <(grep -vE '^[[:space:]]*#' "$CI_YML" \
-    | grep -oE 'packages/mosaic/framework/tools/[A-Za-z0-9_./-]+\.(sh|py)' | sort -u)
+    | grep -oE 'packages/mosaic/framework/[A-Za-z0-9_./-]+\.(sh|py)' | sort -u)
 
 # --- Union, and its population-restricted view -------------------------------
 declare -A ENUM=() ENUM_POP=()
@@ -154,7 +160,7 @@ while IFS= read -r f; do
         fail "UNENUMERATED: '$rel' exists on disk but is neither enumerated on any CI surface nor signed in the exclusions file"
         unlisted=$(( unlisted + 1 ))
     fi
-done < <(find "$TOOLS_DIR" -type f -name '*.sh' | sort)
+done < <(find "$FRAMEWORK_DIR" -type f -name '*.sh' | sort)
 
 if (( fail_count > 0 )); then
     printf 'enumeration guard: %d failure(s) — population %d, enumerated (in-population) %d, excluded %d\n' \

packages/mosaic/framework/tools/quality/scripts/test-check-test-enumeration.sh

@@ -161,6 +161,34 @@ R="$(fixture n7)"
 excl "$R" "packages/mosaic/framework/tools/quality/scripts/verify-thing.sh | not a suite but signing it anyway"
 expect NEEDLE 1 "out-of-population exclusion rejected" --out "EXCLUSION OUTSIDE POPULATION" -- "$R"
 
+echo "=== n9/c5: a suite in a SIBLING of tools/ is in the population (@scooby, 2026-08-16) ==="
+# Every other fixture here lives under framework/tools/, which is how the guard came to
+# scan TOOLS_DIR while claiming a population defined by basename alone. The real specimen
+# was framework/systemd/user/test-fleet-units.sh: a member by the guard's own definition,
+# on no CI surface, in no exclusion, and structurally unreachable by the scan. n9 is that
+# blind spot; without it a future narrowing back to TOOLS_DIR passes all fourteen needles.
+R="$(fixture n9)"
+mkdir -p "$R/packages/mosaic/framework/systemd/user"
+printf '#!/usr/bin/env bash\nexit 0\n' > "$R/packages/mosaic/framework/systemd/user/test-sibling.sh"
+expect NEEDLE 1 "suite outside tools/ but inside framework/ is enumerable, not invisible" \
+    --out "UNENUMERATED: 'packages/mosaic/framework/systemd/user/test-sibling.sh'" -- "$R"
+# c5 is why the S2 regex had to widen WITH the scan: detecting the file is useless if the
+# fix for it cannot be recognised. Enumerating a sibling-directory suite on ci.yml must
+# clear the finding — under a tools/-scoped S2 it stays UNENUMERATED forever and the only
+# reachable disposition is an exclusion.
+#
+# Measured scope of what c5 catches, because it is narrower than it looks: against the
+# ORIGINAL guard (both hunks absent) c5 passes vacuously — the scan never sees the file
+# and S2 never matches it, so nothing is asserted. It discriminates against the HALF-patch
+# — scan widened, S2 narrowed back — which is the realistic future regression, and it was
+# confirmed red in exactly that state. n9 is the one that fails on the original.
+R="$(fixture c5)"
+mkdir -p "$R/packages/mosaic/framework/systemd/user"
+printf '#!/usr/bin/env bash\nexit 0\n' > "$R/packages/mosaic/framework/systemd/user/test-sibling.sh"
+printf '      - bash packages/mosaic/framework/systemd/user/test-sibling.sh\n' >> "$R/.woodpecker/ci.yml"
+expect CONTROL 0 "enumerating a sibling-directory suite on ci.yml actually clears it" \
+    --out "enumeration guard: OK" -- "$R"
+
 echo
 printf 'enumeration-guard needles: %d passed, %d failed\n' "$PASS" "$FAIL"
 (( FAIL == 0 ))