@muggleai/works 5.12.0-staging.84 → 5.12.0-staging.86

package/dist/plugin/scripts/guardrail-debug-path-gate.sh

@@ -32,7 +32,8 @@ state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"failedRuns"' "$state_file" \
   || grep -q '"failedRuns": \[\]' "$state_file" \
-  || grep -q '"debugReleased": true' "$state_file" \n  || grep -q '"debugSkipped": true' "$state_file"; then
+  || grep -q '"debugReleased": true' "$state_file" \
+  || grep -q '"debugSkipped": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/dist/plugin/scripts/guardrail-e2e-gate.sh

@@ -29,7 +29,8 @@ fi
 state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"unitTestsGreen": true' "$state_file" \
-  || grep -q '"e2eReleased": true' "$state_file" \n  || grep -q '"e2eRun": true' "$state_file"; then
+  || grep -q '"e2eReleased": true' "$state_file" \
+  || grep -q '"e2eRun": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/dist/plugin/scripts/guardrail-stage-gate.sh

@@ -34,7 +34,8 @@ state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"mandatoryStages"' "$state_file" \
   || grep -q '"mandatoryStages": \[\]' "$state_file" \
-  || grep -q '"stageReleased": true' "$state_file" \n  || grep -q '"stageSkipped": true' "$state_file"; then
+  || grep -q '"stageReleased": true' "$state_file" \
+  || grep -q '"stageSkipped": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/dist/plugin/scripts/guardrail-walkthrough-gate.sh

@@ -32,7 +32,8 @@ state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"e2eRun": true' "$state_file" \
   || grep -q '"walkthroughPosted": true' "$state_file" \
-  || grep -q '"walkthroughReleased": true' "$state_file" \n  || grep -q '"walkthroughSkipped": true' "$state_file"; then
+  || grep -q '"walkthroughReleased": true' "$state_file" \
+  || grep -q '"walkthroughSkipped": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/dist/plugin/scripts/guardrail-watch-gate.sh

@@ -31,7 +31,8 @@ state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"prsHandled"' "$state_file" \
   || grep -q '"prsHandled": \[\]' "$state_file" \
-  || grep -q '"watchReleased": true' "$state_file" \n  || grep -q '"watchSkipped": true' "$state_file"; then
+  || grep -q '"watchReleased": true' "$state_file" \
+  || grep -q '"watchSkipped": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/dist/plugin/scripts/pr-watch-guards.sh

@@ -30,6 +30,15 @@ MUGGLE_PR_WATCH_POLL_INTERVAL="${MUGGLE_PR_WATCH_POLL_INTERVAL:-60}"
 # revoked auth) exhausts it.
 MUGGLE_PR_WATCH_MAX_FETCH_FAILURES="${MUGGLE_PR_WATCH_MAX_FETCH_FAILURES:-60}"
 
+# Seconds a loop will wait for the arming session to write watch-watermark.env
+# before giving up. The loop cannot evaluate a single wake condition without it,
+# so an unseeded slot is a watcher that will never report anything — and one that
+# looks perfectly healthy while it does nothing, because it still holds its PID
+# lease and still touches its heartbeat. Arming writes the file in the same turn
+# it starts the loop, so anything past a couple of minutes means the arming
+# sequence was not followed and the watch is inert.
+MUGGLE_PR_WATCH_MAX_UNSEEDED="${MUGGLE_PR_WATCH_MAX_UNSEEDED:-180}"
+
 # Seconds to sleep after `fails` consecutive failed fetches: the poll interval,
 # then a linear back-off capped at 5 minutes so a sustained outage is retried
 # calmly rather than hammered every 60s.
@@ -59,6 +68,15 @@ watcher_lifetime_exceeded() {
     [ $((now - started)) -ge "$max" ]
 }
 
+# True when the slot has gone unseeded longer than the cap allows. 0 is
+# unbounded, matching `watcher_lifetime_exceeded`, for a caller that seeds the
+# watermark out of band.
+watcher_unseeded_too_long() {
+    local waited="$1" max="${2:-$MUGGLE_PR_WATCH_MAX_UNSEEDED}"
+    [ "$max" -eq 0 ] 2>/dev/null && return 1
+    [ "$waited" -ge "$max" ]
+}
+
 # True when pid names a running process. `kill -0` sends no signal; EPERM means
 # the process exists but is foreign, which still counts as alive. Used by
 # arm-watcher's pre-arm dedup to decide whether a watcher already owns the slot.

package/dist/plugin/scripts/pr-watch-loop.sh

@@ -54,6 +54,7 @@ state_projection="$(cat "${script_dir}/pr-watch-state.jq")"
 echo "$$" > "${slot}/watch.pid"
 started=$(date +%s)
 fails=0
+unseeded=0
 
 # In-memory floors, above the on-disk watermark. The watermark is advanced by
 # the session after it handles a wave; these stop the loop re-reporting an event
@@ -137,11 +138,24 @@ while :; do
     touch "${slot}/watch-heartbeat" 2>/dev/null
 
     # No watermark yet means the arming session has not finished seeding. Wait
-    # rather than treat every floor as zero, which would fire the whole backlog.
+    # rather than treat every floor as zero, which would fire the whole backlog —
+    # but not forever. Every wake condition below is evaluated against a floor
+    # read from this file, so a loop that never gets one polls nothing, reports
+    # nothing and never reaches the terminal check, while still holding its PID
+    # lease and touching its heartbeat. That combination is the worst kind of
+    # broken: reconcile reads the live lease and fresh beacon as healthy and
+    # declines to re-arm, so the PR is silently unwatched for as long as the
+    # session lives. Fail loudly instead.
     if [ ! -f "${slot}/watch-watermark.env" ]; then
+        unseeded=$((unseeded + MUGGLE_PR_WATCH_POLL_INTERVAL))
+        if watcher_unseeded_too_long "$unseeded"; then
+            echo "WATCH-FAIL pr=$pr_number no watch-watermark.env after ${unseeded}s — arming never seeded it, so this watch can detect nothing (arm-watcher.md steps 1-2)"
+            exit 1
+        fi
         sleep "$MUGGLE_PR_WATCH_POLL_INTERVAL"
         continue
     fi
+    unseeded=0
 
     watermark_review=$(read_watermark_value REV)
     watermark_comment=$(read_watermark_value COM)

package/dist/release-manifest.json

@@ -1,7 +1,7 @@
 {
   "release": "5.11.0",
-  "buildId": "run-84-1",
-  "commitSha": "936a62e2fe1a6861faf2f4eb10643d9e00bf8f6f",
-  "buildTime": "2026-08-29T05:05:31Z",
+  "buildId": "run-86-1",
+  "commitSha": "795c9069663dd1ea34e9f2e58ac31ba832ec4137",
+  "buildTime": "2026-08-30T05:45:55Z",
   "serviceName": "muggle-ai-works-mcp"
 }

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@muggleai/works",
   "mcpName": "io.github.multiplex-ai/muggle",
-  "version": "5.12.0-staging.84",
+  "version": "5.12.0-staging.86",
   "description": "Ship quality products with AI-powered E2E acceptance testing that validates your web app like a real user — from Claude Code and Cursor to PR.",
   "type": "module",
   "main": "dist/index.js",

package/plugin/scripts/guardrail-debug-path-gate.sh

@@ -32,7 +32,8 @@ state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"failedRuns"' "$state_file" \
   || grep -q '"failedRuns": \[\]' "$state_file" \
-  || grep -q '"debugReleased": true' "$state_file" \n  || grep -q '"debugSkipped": true' "$state_file"; then
+  || grep -q '"debugReleased": true' "$state_file" \
+  || grep -q '"debugSkipped": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/plugin/scripts/guardrail-e2e-gate.sh

@@ -29,7 +29,8 @@ fi
 state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"unitTestsGreen": true' "$state_file" \
-  || grep -q '"e2eReleased": true' "$state_file" \n  || grep -q '"e2eRun": true' "$state_file"; then
+  || grep -q '"e2eReleased": true' "$state_file" \
+  || grep -q '"e2eRun": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/plugin/scripts/guardrail-stage-gate.sh

@@ -34,7 +34,8 @@ state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"mandatoryStages"' "$state_file" \
   || grep -q '"mandatoryStages": \[\]' "$state_file" \
-  || grep -q '"stageReleased": true' "$state_file" \n  || grep -q '"stageSkipped": true' "$state_file"; then
+  || grep -q '"stageReleased": true' "$state_file" \
+  || grep -q '"stageSkipped": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/plugin/scripts/guardrail-walkthrough-gate.sh

@@ -32,7 +32,8 @@ state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"e2eRun": true' "$state_file" \
   || grep -q '"walkthroughPosted": true' "$state_file" \
-  || grep -q '"walkthroughReleased": true' "$state_file" \n  || grep -q '"walkthroughSkipped": true' "$state_file"; then
+  || grep -q '"walkthroughReleased": true' "$state_file" \
+  || grep -q '"walkthroughSkipped": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/plugin/scripts/guardrail-watch-gate.sh

@@ -31,7 +31,8 @@ state_file="$home/.muggle-ai/guardrails/$sid.json"
 if [ ! -f "$state_file" ] \
   || ! grep -q '"prsHandled"' "$state_file" \
   || grep -q '"prsHandled": \[\]' "$state_file" \
-  || grep -q '"watchReleased": true' "$state_file" \n  || grep -q '"watchSkipped": true' "$state_file"; then
+  || grep -q '"watchReleased": true' "$state_file" \
+  || grep -q '"watchSkipped": true' "$state_file"; then
   printf '{}'
   exit 0
 fi

package/plugin/scripts/pr-watch-guards.sh

@@ -30,6 +30,15 @@ MUGGLE_PR_WATCH_POLL_INTERVAL="${MUGGLE_PR_WATCH_POLL_INTERVAL:-60}"
 # revoked auth) exhausts it.
 MUGGLE_PR_WATCH_MAX_FETCH_FAILURES="${MUGGLE_PR_WATCH_MAX_FETCH_FAILURES:-60}"
 
+# Seconds a loop will wait for the arming session to write watch-watermark.env
+# before giving up. The loop cannot evaluate a single wake condition without it,
+# so an unseeded slot is a watcher that will never report anything — and one that
+# looks perfectly healthy while it does nothing, because it still holds its PID
+# lease and still touches its heartbeat. Arming writes the file in the same turn
+# it starts the loop, so anything past a couple of minutes means the arming
+# sequence was not followed and the watch is inert.
+MUGGLE_PR_WATCH_MAX_UNSEEDED="${MUGGLE_PR_WATCH_MAX_UNSEEDED:-180}"
+
 # Seconds to sleep after `fails` consecutive failed fetches: the poll interval,
 # then a linear back-off capped at 5 minutes so a sustained outage is retried
 # calmly rather than hammered every 60s.
@@ -59,6 +68,15 @@ watcher_lifetime_exceeded() {
     [ $((now - started)) -ge "$max" ]
 }
 
+# True when the slot has gone unseeded longer than the cap allows. 0 is
+# unbounded, matching `watcher_lifetime_exceeded`, for a caller that seeds the
+# watermark out of band.
+watcher_unseeded_too_long() {
+    local waited="$1" max="${2:-$MUGGLE_PR_WATCH_MAX_UNSEEDED}"
+    [ "$max" -eq 0 ] 2>/dev/null && return 1
+    [ "$waited" -ge "$max" ]
+}
+
 # True when pid names a running process. `kill -0` sends no signal; EPERM means
 # the process exists but is foreign, which still counts as alive. Used by
 # arm-watcher's pre-arm dedup to decide whether a watcher already owns the slot.

package/plugin/scripts/pr-watch-loop.sh

@@ -54,6 +54,7 @@ state_projection="$(cat "${script_dir}/pr-watch-state.jq")"
 echo "$$" > "${slot}/watch.pid"
 started=$(date +%s)
 fails=0
+unseeded=0
 
 # In-memory floors, above the on-disk watermark. The watermark is advanced by
 # the session after it handles a wave; these stop the loop re-reporting an event
@@ -137,11 +138,24 @@ while :; do
     touch "${slot}/watch-heartbeat" 2>/dev/null
 
     # No watermark yet means the arming session has not finished seeding. Wait
-    # rather than treat every floor as zero, which would fire the whole backlog.
+    # rather than treat every floor as zero, which would fire the whole backlog —
+    # but not forever. Every wake condition below is evaluated against a floor
+    # read from this file, so a loop that never gets one polls nothing, reports
+    # nothing and never reaches the terminal check, while still holding its PID
+    # lease and touching its heartbeat. That combination is the worst kind of
+    # broken: reconcile reads the live lease and fresh beacon as healthy and
+    # declines to re-arm, so the PR is silently unwatched for as long as the
+    # session lives. Fail loudly instead.
     if [ ! -f "${slot}/watch-watermark.env" ]; then
+        unseeded=$((unseeded + MUGGLE_PR_WATCH_POLL_INTERVAL))
+        if watcher_unseeded_too_long "$unseeded"; then
+            echo "WATCH-FAIL pr=$pr_number no watch-watermark.env after ${unseeded}s — arming never seeded it, so this watch can detect nothing (arm-watcher.md steps 1-2)"
+            exit 1
+        fi
         sleep "$MUGGLE_PR_WATCH_POLL_INTERVAL"
         continue
     fi
+    unseeded=0
 
     watermark_review=$(read_watermark_value REV)
     watermark_comment=$(read_watermark_value COM)