[ci] Get the fork's main green: fix the four standing CI failures

.github/workflows/ci.yml

@@ -439,8 +439,8 @@ jobs:
           snapshot_output=$(/bin/bash tests/fm-fleet-snapshot-view.test.sh)
           printf '%s\n' "$snapshot_output"
           snapshot_count=$(printf '%s\n' "$snapshot_output" | grep -c '^ok - ')
-          [ "$snapshot_count" -eq 18 ] || {
-            echo "::error::expected 18 snapshot/fleet-view tests, got $snapshot_count"
+          [ "$snapshot_count" -eq 19 ] || {
+            echo "::error::expected 19 snapshot/fleet-view tests, got $snapshot_count"
             exit 1
           }
 

docs/configuration.md

@@ -512,7 +512,8 @@ An absent or unknown named row also makes the candidate unrankable and is report
 `ultra` is native-only: the model-aware validation contract and launch mapping are owned by `bin/fm-harness.sh validate-native-effort` and `bin/fm-spawn.sh` respectively.
 Codex `max` is valid when the profile selects `gpt-5.6-luna`, whose installed catalog entry supports that reasoning level.
 An omitted model or effort means the selected harness uses its own default for that axis.
-A `cline` profile's `model` is the full `<provider>/<model>` id cline expects (for example `cline-pass/deepseek-v4-flash` or `cline-pass/glm-5.3`); cline derives the provider from that prefix, so no separate provider field is needed.
+A `cline` profile's `model` is the full `<provider>/<model>` id cline expects (for example `cline-pass/deepseek-v4-flash` or `cline-pass/glm-5.3`), and cline derives its launch provider from that prefix.
+Typed resolution still needs the profile's `provider` (`cline-pass` for a ClinePass model), because cline is not in the single-provider table above and the resolver's `provider` names the quota-axi provider family rather than the launch prefix.
 Every profile array is an implicit quota-aware choice resolved through `quota-array-dispatch`.
 If no dispatch rule fits, firstmate resolves `default` through the same object-or-array path before falling back to `config/crew-harness`.
 Except for `ultra`, which refuses unsupported profiles under the native-effort contract above, an effort value the chosen harness does not accept is recorded as `effort=` in task meta for traceability but omitted from the launch flags.

docs/examples/crew-dispatch.json

@@ -15,24 +15,24 @@
       "use": [
         { "harness": "claude", "model": "claude-sonnet-5", "effort": "high" },
         { "harness": "codex", "model": "gpt-5.5", "effort": "high" },
-        { "harness": "cline", "model": "cline-pass/deepseek-v4-pro", "effort": "high" }
+        { "harness": "cline", "model": "cline-pass/deepseek-v4-pro", "effort": "high", "provider": "cline-pass" }
       ],
       "why": "Use a strong coding profile for big, ambiguous work; resolve the alternatives through quota-array-dispatch. The ClinePass open-weights seat keeps a parallel lane open when a subscription quota dies."
     },
     {
       "when": "The task is a medium-sized coding ship that can run in parallel with others and a ClinePass open-weights coding model is a cheaper seat than a frontier subscription.",
       "use": [
-        { "harness": "cline", "model": "cline-pass/glm-5.3", "effort": "high" },
-        { "harness": "cline", "model": "cline-pass/kimi-k2.7-code", "effort": "high" },
-        { "harness": "cline", "model": "cline-pass/deepseek-v4-flash", "effort": "xhigh" }
+        { "harness": "cline", "model": "cline-pass/glm-5.3", "effort": "high", "provider": "cline-pass" },
+        { "harness": "cline", "model": "cline-pass/kimi-k2.7-code", "effort": "high", "provider": "cline-pass" },
+        { "harness": "cline", "model": "cline-pass/deepseek-v4-flash", "effort": "xhigh", "provider": "cline-pass" }
       ],
       "why": "Use the ClinePass pool as a third coding lane for medium ships so a scarce frontier or Gemini seat is not overloaded."
     }
   ],
   "default": [
     { "harness": "codex", "model": "gpt-5.5", "effort": "medium" },
     { "harness": "pi", "model": "anthropic/claude-sonnet-5", "effort": "medium", "provider": "claude" },
-    { "harness": "cline", "model": "cline-pass/deepseek-v4-flash", "effort": "high" }
+    { "harness": "cline", "model": "cline-pass/deepseek-v4-flash", "effort": "high", "provider": "cline-pass" }
   ],
   "providerCaps": { "default": 4 }
 }

tests/fm-claim.test.sh

@@ -221,9 +221,11 @@ case "$OUT" in
 *) fail "list did not include the recorded claim: '$OUT'" ;;
 esac
 
-# A corrupt record fails closed rather than being trusted.
-CORRUPT=$(find "$FM_CLAIM_ROOT" -name '*.claim' | head -n 1)
-[ -n "$CORRUPT" ] || fail "no claim file was created to corrupt"
+# A corrupt record fails closed rather than being trusted. Three claims exist
+# here, so pick the one whose documented key= line names this target instead
+# of trusting directory order, which differs between filesystems.
+CORRUPT=$(grep -l '^key=pr:github.com/owner/repo#11$' "$FM_CLAIM_ROOT"/*.claim 2>/dev/null | head -n 1)
+[ -n "$CORRUPT" ] || fail "no claim record for owner/repo#11 was found to corrupt"
 printf 'garbage\n' >"$CORRUPT"
 run "$CLAIM" status "owner/repo#11"
 assert_rc 5

tests/fm-dispatch-resolve.test.sh

@@ -320,7 +320,7 @@ assert_contains "$out" "  profile: --harness 'gemini' --model 'gemini-3.8-flash-
 
 cp "$ROOT/docs/examples/crew-dispatch.json" "$RULES"
 cat > "$RESPONSE" <<'JSON'
-{"model":"jev-1.13.0","answers":{"rule":{"type":"choice","choice":"default","confidence":0.9,"probabilities":{"rule_1":0.02,"rule_2":0.02,"rule_3":0.02,"default":0.94}}},"usage":{"input_tokens":812,"output_tokens":60}}
+{"model":"jev-1.13.0","answers":{"rule":{"type":"choice","choice":"default","confidence":0.9,"probabilities":{"rule_1":0.02,"rule_2":0.02,"rule_3":0.02,"rule_4":0.02,"default":0.92}}},"usage":{"input_tokens":812,"output_tokens":60}}
 JSON
 reset_log
 TYPESAFE_API_KEY=$KEY run code out err "$BRIEF"

tests/fm-home-summary-refresh.test.sh

@@ -839,8 +839,14 @@ while [ ! -e "$RESTART_LOCK_MARKER" ] && [ "$i" -lt 100 ]; do
   i=$((i + 1))
 done
 [ -e "$RESTART_LOCK_MARKER" ] || fail "could not hold the publication lock for restart coverage"
+# FM_HOME_SUMMARY_TIMEOUT below bounds the watcher's own detached refresh,
+# which fires on every poll while the ledger is absent and races the idle-only
+# refresh at the end of this section for the dead lock. It must comfortably
+# exceed one producer run on a slow CI runner: a killed attempt publishes
+# nothing and releases the lock, the next poll's attempt dies the same way,
+# and the ledger never appears.
 PATH="$FAKEBIN:$PATH" FM_ROOT_OVERRIDE="$ROOT" FM_HOME="$RESTART_HOME" \
-  FM_POLL=1 FM_HOME_SUMMARY_INTERVAL=999999 FM_HOME_SUMMARY_TIMEOUT=2 \
+  FM_POLL=1 FM_HOME_SUMMARY_INTERVAL=999999 FM_HOME_SUMMARY_TIMEOUT=30 \
   FM_SIGNAL_GRACE=0 FM_CHECK_INTERVAL=9999999 FM_HEARTBEAT=9999999 \
   "$WATCH" > "$TMP_ROOT/restart-watch-one.out" 2> "$TMP_ROOT/restart-watch-one.err" &
 WATCH_PID=$!
@@ -865,7 +871,7 @@ wait "$WATCH_PID" >/dev/null 2>&1 || true
 WATCH_PID=
 rm -f "$RESTART_HOME/state/.last-watcher-beat"
 PATH="$FAKEBIN:$PATH" FM_ROOT_OVERRIDE="$ROOT" FM_HOME="$RESTART_HOME" \
-  FM_POLL=1 FM_HOME_SUMMARY_INTERVAL=999999 FM_HOME_SUMMARY_TIMEOUT=2 \
+  FM_POLL=1 FM_HOME_SUMMARY_INTERVAL=999999 FM_HOME_SUMMARY_TIMEOUT=30 \
   FM_SIGNAL_GRACE=0 FM_CHECK_INTERVAL=9999999 FM_HEARTBEAT=9999999 \
   "$WATCH" > "$TMP_ROOT/restart-watch-two.out" 2> "$TMP_ROOT/restart-watch-two.err" &
 WATCH_PID=$!
@@ -884,7 +890,7 @@ if ! kill -0 "$WATCH_PID" 2>/dev/null; then
   wait "$WATCH_PID" >/dev/null 2>&1 || true
   rm -f "$RESTART_HOME/state/.last-watcher-beat"
   PATH="$FAKEBIN:$PATH" FM_ROOT_OVERRIDE="$ROOT" FM_HOME="$RESTART_HOME" \
-    FM_POLL=1 FM_HOME_SUMMARY_INTERVAL=999999 FM_HOME_SUMMARY_TIMEOUT=2 \
+    FM_POLL=1 FM_HOME_SUMMARY_INTERVAL=999999 FM_HOME_SUMMARY_TIMEOUT=30 \
     FM_SIGNAL_GRACE=0 FM_CHECK_INTERVAL=9999999 FM_HEARTBEAT=9999999 \
     "$WATCH" > "$TMP_ROOT/restart-watch-three.out" 2> "$TMP_ROOT/restart-watch-three.err" &
   WATCH_PID=$!