fix(harness): count what the roster run ADDED, not what the file holds

Forcing the planner onto the local link produced a green chain and a red
assertion:

  roster: the planner sized this mission at 1 member(s)          PASS
  roster: ROSTER.md has 3 line(s) for a 1-member roster          FAIL

The model was right and the check was wrong. ROSTER.md does not start empty —
the auto-merge work put an earlier run's two lines onto main — so a 1-member
roster that correctly appended one line delivered three, and the scenario
reported a model that had ignored its own proposal.

It now measures the DELTA against main. Any assertion against a scratch repo
that accumulates has to, or it decays into a test of how many times it has been
run before.

Proven on the local model end to end: opus 429 -> local:ornith-fleet:9b
answered -> `mission_roster: ... local:ornith-fleet:9b proposed 1 member(s)` ->
the composed graph ran -> the branch added exactly one line. 5/5.

CLAWMATES_MODEL_FALLBACK is removed from gw-04's .env again; it was set only to
force the last link for this test, and the deployed default is the full chain.

Co-Authored-By: Claude Opus 5 <[email protected]>

scripts/verify-mission-delivery.sh

@@ -252,6 +252,20 @@ for a in d.get("artifacts") or []:
     'https://git.redclaw.dev/api/v1/repos/$repo/raw/$3?ref=$enc'"
 }
 
+# Read a file from the repo's DEFAULT branch, so a scenario can measure what a
+# run ADDED rather than what the file happens to contain. Prints nothing and
+# succeeds when the file is absent — an empty baseline is a real baseline.
+fetch_main() { # fetch_main <path>
+  local token repo
+  token=$(ssh "$HOST" 'docker exec clawmates_server_1 printenv GITEA_TOKEN' | tr -d '\r')
+  repo=$(ssh "$HOST" "docker exec clawmates_postgres_1 psql -U postgres -d clawmates -tAc \
+    \"select owner || '/' || name from repos where id='$REPO_ID';\"" \
+    | head -1 | tr -d '[:space:]')
+  [ -n "$token" ] && [ -n "$repo" ] || return 0
+  ssh "$HOST" "curl -sf -H 'Authorization: token $token' \
+    'https://git.redclaw.dev/api/v1/repos/$repo/raw/$1?ref=main'" 2>/dev/null || true
+}
+
 # ── Scenario: a phase runs inside a microVM, and fans out ─────────
 #
 # Guards everything the microVM track proved by hand: that the agent ran in a
@@ -816,10 +830,19 @@ if d: print(d[0]["id"], len(d[0]["roster"]["members"]))
   delivered=$(fetch_delivered "$token" "$mission" ROSTER.md) \
     || { fail "roster: could not read ROSTER.md from the pushed branch"; return 1; }
   lines=$(printf '%s\n' "$delivered" | grep -c '[^[:space:]]')
-  if [ "${lines:-0}" = "$members" ]; then
+  # Counted as a DELTA against main, not as a total.
-    pass "roster: ROSTER.md has one line per proposed member ($lines)"
+  #
+  # The absolute form assumed ROSTER.md starts empty, and it does not: the
+  # auto-merge work put an earlier run's two lines onto main, so the next
+  # 1-member roster delivered three lines and was reported as a model that
+  # ignored its own proposal. It had done exactly the right thing. Any check
+  # against a scratch repo that ACCUMULATES has to measure what this run added.
+  base=$(fetch_main "ROSTER.md" | grep -c '[^[:space:]]')
+  local added=$((lines - base))
+  if [ "$added" = "$members" ]; then
+    pass "roster: this run added one line per proposed member ($added, on top of $base already on main)"
   else
-    fail "roster: ROSTER.md has $lines line(s) for a $members-member roster: $(printf '%s' "$delivered" | tr '\n' '|')"
+    fail "roster: this run added $added line(s) for a $members-member roster (branch=$lines main=$base): $(printf '%s' "$delivered" | tr '\n' '|')"
   fi
   check_single_uid "$mission" roster
 }