prizmkit 1.1.66 → 1.1.67

package/bundled/VERSION.json

@@ -1,5 +1,5 @@
 {
-  "frameworkVersion": "1.1.66",
-  "bundledAt": "2026-06-08T19:45:48.630Z",
-  "bundledFrom": "940cbd4"
+  "frameworkVersion": "1.1.67",
+  "bundledAt": "2026-06-09T02:37:28.761Z",
+  "bundledFrom": "d4b8c30"
 }

package/bundled/adapters/codex/settings-adapter.js

@@ -20,7 +20,7 @@ project_doc_fallback_filenames = ["CLAUDE.md", "CODEBUDDY.md"]
 
 [agents]
 max_depth = 1
-job_max_runtime_seconds = 840
+job_max_runtime_seconds = 3300
 `;
 
   await writeFile(configPath, configToml);

package/bundled/dev-pipeline/.env.example

@@ -41,6 +41,9 @@
 # ─── Logging & Heartbeat ─────────────────────────────────────────────
 # HEARTBEAT_INTERVAL=30      # Heartbeat log interval in seconds
 # HEARTBEAT_STALE_THRESHOLD=600  # Max seconds without heartbeat before marking stale
+# STALE_KILL_THRESHOLD=900   # Auto-kill after N seconds without parent log progress (0 = disabled)
+# CODEX_WAIT_STALE_KILL_THRESHOLD=3600  # Longer no-log window while Codex waits on subagents
+# CODEX_SUBAGENT_TIMEOUT_SECONDS=3300   # Codex subagent max runtime; defaults to wait threshold - 300
 # LOG_CLEANUP_ENABLED=1      # Periodic log cleanup (1=on, 0=off)
 # LOG_RETENTION_DAYS=14      # Delete logs older than N days
 # LOG_MAX_TOTAL_MB=1024      # Keep total logs under N MB via oldest-first cleanup

package/bundled/dev-pipeline/SCHEMA_ANALYSIS.md

@@ -353,6 +353,9 @@ pending, in_progress, completed, failed, skipped
 | `DEV_BRANCH` | string | auto-generated | Custom branch name |
 | `HEARTBEAT_INTERVAL` | integer | 30 | Heartbeat log interval (s) |
 | `HEARTBEAT_STALE_THRESHOLD` | integer | 600 | Max seconds without heartbeat |
+| `STALE_KILL_THRESHOLD` | integer | 900 | Auto-kill after N seconds without parent log progress |
+| `CODEX_WAIT_STALE_KILL_THRESHOLD` | integer | 3600 | Longer no-log stale window while Codex waits on subagents |
+| `CODEX_SUBAGENT_TIMEOUT_SECONDS` | integer | 3300 | Codex subagent max runtime |
 | `LOG_CLEANUP_ENABLED` | integer | 1 | Periodic cleanup |
 | `LOG_RETENTION_DAYS` | integer | 14 | Delete logs older than N days |
 | `LOG_MAX_TOTAL_MB` | integer | 1024 | Max total logs (MB) |
@@ -532,4 +535,3 @@ Located in `/dev-pipeline/templates/`:
 - Agent: 6 files
 - Base/Shared: 7 files
 - Singleton: 3 files
-

package/bundled/dev-pipeline/lib/common.sh

@@ -344,6 +344,23 @@ prizm_detect_cli_and_platform() {
 # command substitution; the background process must remain a child of the
 # runner shell so wait/heartbeat/trap handling works correctly.
 PRIZM_AI_PID=""
+
+_prizm_codex_subagent_timeout_seconds() {
+    local configured="${CODEX_SUBAGENT_TIMEOUT_SECONDS:-}"
+    if [[ "$configured" =~ ^[0-9]+$ && "$configured" -gt 0 ]]; then
+        printf '%s\n' "$configured"
+        return 0
+    fi
+
+    local wait_threshold="${CODEX_WAIT_STALE_KILL_THRESHOLD:-3600}"
+    if [[ "$wait_threshold" =~ ^[0-9]+$ && "$wait_threshold" -gt 600 ]]; then
+        printf '%s\n' "$((wait_threshold - 300))"
+        return 0
+    fi
+
+    printf '%s\n' 3300
+}
+
 prizm_start_ai_session() {
     local prompt_path="$1"
     local log_path="$2"
@@ -370,15 +387,8 @@ prizm_start_ai_session() {
             ;;
         codex)
             local codex_args=(--ask-for-approval never --sandbox danger-full-access)
-            local codex_subagent_timeout="${CODEX_SUBAGENT_TIMEOUT_SECONDS:-}"
-            if [[ -z "$codex_subagent_timeout" ]]; then
-                local outer_stale_threshold="${STALE_KILL_THRESHOLD:-900}"
-                if [[ "$outer_stale_threshold" =~ ^[0-9]+$ && "$outer_stale_threshold" -gt 120 ]]; then
-                    codex_subagent_timeout=$((outer_stale_threshold - 60))
-                else
-                    codex_subagent_timeout=840
-                fi
-            fi
+            local codex_subagent_timeout
+            codex_subagent_timeout="$(_prizm_codex_subagent_timeout_seconds)"
             if [[ "$codex_subagent_timeout" =~ ^[0-9]+$ && "$codex_subagent_timeout" -gt 0 ]]; then
                 codex_args+=(--config "agents.job_max_runtime_seconds=$codex_subagent_timeout")
             fi
@@ -430,15 +440,8 @@ prizm_run_ai_session() {
             ;;
         codex)
             local codex_args=(--ask-for-approval never --sandbox danger-full-access)
-            local codex_subagent_timeout="${CODEX_SUBAGENT_TIMEOUT_SECONDS:-}"
-            if [[ -z "$codex_subagent_timeout" ]]; then
-                local outer_stale_threshold="${STALE_KILL_THRESHOLD:-900}"
-                if [[ "$outer_stale_threshold" =~ ^[0-9]+$ && "$outer_stale_threshold" -gt 120 ]]; then
-                    codex_subagent_timeout=$((outer_stale_threshold - 60))
-                else
-                    codex_subagent_timeout=840
-                fi
-            fi
+            local codex_subagent_timeout
+            codex_subagent_timeout="$(_prizm_codex_subagent_timeout_seconds)"
             if [[ "$codex_subagent_timeout" =~ ^[0-9]+$ && "$codex_subagent_timeout" -gt 0 ]]; then
                 codex_args+=(--config "agents.job_max_runtime_seconds=$codex_subagent_timeout")
             fi

package/bundled/dev-pipeline/lib/heartbeat.sh

@@ -41,6 +41,7 @@ start_heartbeat() {
     (
         local elapsed=0
         local prev_size=0
+        local prev_child_activity_signature=""
         local stale_seconds=0
         while kill -0 "$cli_pid" 2>/dev/null; do
             sleep "$heartbeat_interval"
@@ -57,8 +58,41 @@ start_heartbeat() {
             local growth=$((cur_size - prev_size))
             prev_size=$cur_size
 
-            # Track progress staleness (no log growth = stale)
-            if [[ $growth -eq 0 ]]; then
+            local child_activity_signature=""
+            local child_total_bytes=0
+            local child_session_count=0
+            if [[ -f "$progress_json" ]]; then
+                local child_activity_data
+                child_activity_data=$(python3 - "$progress_json" <<'PY' 2>/dev/null || true
+import json
+import sys
+
+try:
+    with open(sys.argv[1], "r", encoding="utf-8") as fh:
+        progress = json.load(fh)
+except Exception:
+    sys.exit(0)
+
+signature = str(progress.get("child_activity_signature") or "")
+total_bytes = int(progress.get("child_total_bytes") or 0)
+session_count = len(progress.get("child_session_files") or [])
+print(f"{signature}\t{total_bytes}\t{session_count}")
+PY
+                )
+                if [[ -n "$child_activity_data" ]]; then
+                    IFS=$'\t' read -r child_activity_signature child_total_bytes child_session_count <<< "$child_activity_data"
+                fi
+            fi
+
+            local child_growth=0
+            if [[ -n "$child_activity_signature" && "$child_activity_signature" != "$prev_child_activity_signature" ]]; then
+                child_growth=1
+            fi
+            prev_child_activity_signature="$child_activity_signature"
+
+            # Track progress staleness. A Codex parent can sit in `wait`
+            # while child transcripts keep growing, so child activity counts.
+            if [[ $growth -eq 0 && $child_growth -eq 0 ]]; then
                 stale_seconds=$((stale_seconds + heartbeat_interval))
             else
                 stale_seconds=0
@@ -72,28 +106,87 @@ start_heartbeat() {
             else
                 size_display="${cur_size}B"
             fi
+            local child_display=""
+            if [[ ${child_total_bytes:-0} -gt 0 ]]; then
+                local child_size_display
+                if [[ $child_total_bytes -gt 1048576 ]]; then
+                    child_size_display="$((child_total_bytes / 1048576))MB"
+                elif [[ $child_total_bytes -gt 1024 ]]; then
+                    child_size_display="$((child_total_bytes / 1024))KB"
+                else
+                    child_size_display="${child_total_bytes}B"
+                fi
+                child_display=" | child: ${child_size_display}"
+                if [[ ${child_session_count:-0} -gt 1 ]]; then
+                    child_display="${child_display}/${child_session_count}"
+                fi
+            fi
 
             local mins=$((elapsed / 60))
             local secs=$((elapsed % 60))
 
             local status_icon
-            if [[ $growth -gt 0 ]]; then
+            if [[ $growth -gt 0 || $child_growth -gt 0 ]]; then
                 status_icon="${GREEN}▶${NC}"
             else
                 status_icon="${YELLOW}⏸${NC}"
             fi
 
-            # Stale-kill: auto-terminate process if no progress for too long
-            if [[ $stale_kill_threshold -gt 0 && $stale_seconds -ge $stale_kill_threshold ]]; then
+            local effective_stale_kill_threshold="$stale_kill_threshold"
+            if [[ $stale_kill_threshold -gt 0 && -f "$progress_json" ]]; then
+                local codex_wait_threshold
+                codex_wait_threshold=$(python3 - "$progress_json" "$stale_kill_threshold" <<'PY' 2>/dev/null || true
+import json
+import os
+import sys
+
+progress_path = sys.argv[1]
+base_threshold = int(sys.argv[2])
+
+with open(progress_path, "r", encoding="utf-8") as fh:
+    progress = json.load(fh)
+
+spawn_count = 0
+for tool in progress.get("tool_calls", []):
+    if isinstance(tool, dict) and tool.get("name") == "spawn_agent":
+        try:
+            spawn_count += int(tool.get("count", 0))
+        except (TypeError, ValueError):
+            pass
+
+if (
+    progress.get("event_format") == "codex-json"
+    and progress.get("current_tool") == "wait"
+    and spawn_count > 0
+):
+    configured = os.environ.get("CODEX_WAIT_STALE_KILL_THRESHOLD", "")
+    try:
+        wait_threshold = int(configured)
+    except ValueError:
+        wait_threshold = max(base_threshold * 4, 3600)
+    if wait_threshold > base_threshold:
+        print(wait_threshold)
+PY
+                )
+                if [[ "$codex_wait_threshold" =~ ^[0-9]+$ && "$codex_wait_threshold" -gt "$stale_kill_threshold" ]]; then
+                    effective_stale_kill_threshold="$codex_wait_threshold"
+                fi
+            fi
+
+            # Stale-kill: auto-terminate process if no progress for too long.
+            # Codex parent sessions can sit on the `wait` tool while a spawned
+            # subagent is still doing useful work. Give that valid wait a longer
+            # stale window; normal single-agent stalls still use the base limit.
+            if [[ $effective_stale_kill_threshold -gt 0 && $stale_seconds -ge $effective_stale_kill_threshold ]]; then
                 local stale_mins=$((stale_seconds / 60))
-                echo -e "  ${RED}[HEARTBEAT]${NC} ${mins}m${secs}s | log: ${size_display} | ${RED}STALE-KILL: no progress for ${stale_mins}m (threshold: ${stale_kill_threshold}s)${NC}"
+                echo -e "  ${RED}[HEARTBEAT]${NC} ${mins}m${secs}s | log: ${size_display} | ${RED}STALE-KILL: no progress for ${stale_mins}m (threshold: ${effective_stale_kill_threshold}s)${NC}"
                 echo -e "  ${RED}[HEARTBEAT]${NC} Killing AI CLI process $cli_pid (stale session)..."
                 # Write the marker before killing. Some CLIs exit quickly, and the
                 # parent runner may stop this heartbeat process immediately after
                 # wait(1) returns.
                 local _marker_dir
                 _marker_dir="$(dirname "$session_log")"
-                echo "{\"killed_at\": \"$(date -u +%Y-%m-%dT%H:%M:%SZ)\", \"reason\": \"stale_session\", \"stale_seconds\": $stale_seconds, \"threshold\": $stale_kill_threshold}" > "$_marker_dir/stale-kill.json" 2>/dev/null || true
+                echo "{\"killed_at\": \"$(date -u +%Y-%m-%dT%H:%M:%SZ)\", \"reason\": \"stale_session\", \"stale_seconds\": $stale_seconds, \"threshold\": $effective_stale_kill_threshold}" > "$_marker_dir/stale-kill.json" 2>/dev/null || true
                 kill -TERM "$cli_pid" 2>/dev/null || true
                 # Give process 10s to exit gracefully, then force kill
                 local stale_kill_grace_seconds="${STALE_KILL_GRACE_SECONDS:-10}"
@@ -109,9 +202,9 @@ start_heartbeat() {
 
             # Build staleness hint for display
             local stale_hint=""
-            if [[ $stale_kill_threshold -gt 0 && $stale_seconds -gt 0 ]]; then
+            if [[ $effective_stale_kill_threshold -gt 0 && $stale_seconds -gt 0 ]]; then
                 local stale_mins=$((stale_seconds / 60))
-                local threshold_mins=$((stale_kill_threshold / 60))
+                local threshold_mins=$((effective_stale_kill_threshold / 60))
                 stale_hint=" | stale: ${stale_mins}m/${threshold_mins}m"
             fi
 
@@ -134,7 +227,7 @@ try:
 except Exception:
     sys.exit(1)
 " "$progress_json" 2>/dev/null) && {
-                    echo -e "  ${status_icon} ${BLUE}[HEARTBEAT]${NC} ${mins}m${secs}s | log: ${size_display} | ${phase}${stale_hint}"
+                    echo -e "  ${status_icon} ${BLUE}[HEARTBEAT]${NC} ${mins}m${secs}s | log: ${size_display}${child_display} | ${phase}${stale_hint}"
                     continue
                 }
             fi
@@ -145,7 +238,7 @@ except Exception:
                 last_activity=$(tail -20 "$session_log" 2>/dev/null | grep -v '^$' | tail -1 | cut -c1-80 || echo "")
             fi
 
-            echo -e "  ${status_icon} ${BLUE}[HEARTBEAT]${NC} ${mins}m${secs}s elapsed | log: ${size_display} (+${growth}B) | ${last_activity}${stale_hint}"
+            echo -e "  ${status_icon} ${BLUE}[HEARTBEAT]${NC} ${mins}m${secs}s elapsed | log: ${size_display}${child_display} (+${growth}B) | ${last_activity}${stale_hint}"
         done
     ) &
     _HEARTBEAT_PID=$!

package/bundled/dev-pipeline/scripts/parse-stream-progress.py

@@ -23,6 +23,7 @@ import tempfile
 import time
 from collections import Counter
 from datetime import datetime, timezone
+from pathlib import Path
 
 
 # Ordered pipeline phases — index defines forward-only progression.
@@ -76,6 +77,13 @@ class ProgressTracker:
         self.event_format = ""
         self.active_subagent_count = 0
         self.subagent_status_counts = Counter()
+        self.codex_child_thread_ids = set()
+        self.child_session_files = []
+        self.child_total_bytes = 0
+        self.child_activity_signature = ""
+        self.last_child_activity_at = ""
+        self._codex_child_session_paths = {}
+        self._last_child_scan_at = 0.0
         self._text_buffer = ""
         self._in_tool_use = False
         self._current_tool_input_parts = []
@@ -113,6 +121,9 @@ class ProgressTracker:
 
                 elif item_type == "collab_tool_call":
                     tool_name = item.get("tool", "collab")
+                    self._record_codex_child_thread_ids(
+                        item.get("receiver_thread_ids")
+                    )
                     if event_type == "item.started":
                         self.current_tool = tool_name
                         self.tool_call_counts[tool_name] += 1
@@ -345,8 +356,117 @@ class ProgressTracker:
         self.subagent_status_counts = counts
         self.active_subagent_count = active
 
+    def _record_codex_child_thread_ids(self, thread_ids):
+        """Remember Codex child thread IDs reported by collab tool calls."""
+        if not isinstance(thread_ids, list):
+            return
+        for thread_id in thread_ids:
+            if isinstance(thread_id, str) and thread_id.strip():
+                self.codex_child_thread_ids.add(thread_id.strip())
+
+    def _codex_sessions_dir(self):
+        """Return the Codex sessions directory for the current environment."""
+        codex_home = os.environ.get("CODEX_HOME")
+        if codex_home:
+            return Path(codex_home).expanduser() / "sessions"
+        return Path.home() / ".codex" / "sessions"
+
+    def _find_codex_child_session_file(self, thread_id):
+        """Find a Codex transcript file for a child thread ID."""
+        sessions_dir = self._codex_sessions_dir()
+        if not sessions_dir.exists():
+            return None
+
+        try:
+            matches = list(sessions_dir.rglob(f"*{thread_id}.jsonl"))
+        except OSError:
+            return None
+
+        if not matches:
+            return None
+
+        try:
+            matches.sort(key=lambda path: path.stat().st_mtime, reverse=True)
+        except OSError:
+            pass
+        return str(matches[0])
+
+    def refresh_child_session_activity(self, force=False):
+        """Refresh Codex child transcript file stats.
+
+        The heartbeat monitor uses this activity signature to treat subagent
+        transcript growth as real progress while the parent Codex session is
+        blocked in `wait`.
+        """
+        previous_signature = self.child_activity_signature
+
+        if not self.codex_child_thread_ids:
+            self.child_session_files = []
+            self.child_total_bytes = 0
+            self.child_activity_signature = ""
+            self.last_child_activity_at = ""
+            return previous_signature != self.child_activity_signature
+
+        now = time.monotonic()
+        should_scan = (
+            force
+            or self._last_child_scan_at == 0.0
+            or (now - self._last_child_scan_at >= 2.0)
+        )
+        if should_scan:
+            for thread_id in sorted(self.codex_child_thread_ids):
+                path = self._codex_child_session_paths.get(thread_id)
+                if not path or not os.path.exists(path):
+                    found = self._find_codex_child_session_file(thread_id)
+                    if found:
+                        self._codex_child_session_paths[thread_id] = found
+            self._last_child_scan_at = now
+
+        files = []
+        signature_parts = []
+        total_bytes = 0
+        max_mtime = 0.0
+
+        for thread_id in sorted(self.codex_child_thread_ids):
+            path = self._codex_child_session_paths.get(thread_id)
+            if not path:
+                continue
+            try:
+                stat = os.stat(path)
+            except OSError:
+                continue
+
+            total_bytes += stat.st_size
+            max_mtime = max(max_mtime, stat.st_mtime)
+            signature_parts.append(
+                f"{thread_id}:{stat.st_size}:{getattr(stat, 'st_mtime_ns', int(stat.st_mtime * 1_000_000_000))}"
+            )
+            files.append(
+                {
+                    "thread_id": thread_id,
+                    "path": path,
+                    "size": stat.st_size,
+                    "mtime": datetime.fromtimestamp(
+                        stat.st_mtime, timezone.utc
+                    ).strftime("%Y-%m-%dT%H:%M:%SZ"),
+                }
+            )
+
+        self.child_session_files = files
+        self.child_total_bytes = total_bytes
+        self.child_activity_signature = "|".join(signature_parts)
+        self.last_child_activity_at = (
+            datetime.fromtimestamp(max_mtime, timezone.utc).strftime(
+                "%Y-%m-%dT%H:%M:%SZ"
+            )
+            if max_mtime
+            else ""
+        )
+        return previous_signature != self.child_activity_signature
+
     def to_dict(self):
         """Export current state as a dictionary for JSON serialization."""
+        self.refresh_child_session_activity()
         tool_calls = [
             {"name": name, "count": count}
             for name, count in self.tool_call_counts.most_common()
@@ -367,6 +487,11 @@ class ProgressTracker:
             "total_tool_calls": self.total_tool_calls,
             "active_subagent_count": self.active_subagent_count,
             "subagent_states": subagent_states,
+            "child_thread_ids": sorted(self.codex_child_thread_ids),
+            "child_session_files": self.child_session_files,
+            "child_total_bytes": self.child_total_bytes,
+            "child_activity_signature": self.child_activity_signature,
+            "last_child_activity_at": self.last_child_activity_at,
             "last_text_snippet": self.last_text_snippet,
             "is_active": self.is_active,
             "errors": self.errors[-10:],  # Keep last 10 errors
@@ -397,6 +522,15 @@ def tail_and_parse(session_log, progress_file, poll_interval=0.5):
     tracker = ProgressTracker()
     last_write_state = None
 
+    def state_key(state):
+        return (
+            state["message_count"],
+            state["current_tool"],
+            state["current_phase"],
+            state["total_tool_calls"],
+            state.get("child_activity_signature", ""),
+        )
+
     # Wait for log file to appear
     wait_count = 0
     while not os.path.exists(session_log):
@@ -428,22 +562,20 @@ def tail_and_parse(session_log, progress_file, poll_interval=0.5):
 
                 # Write progress if state changed
                 current_state = tracker.to_dict()
-                state_key = (
-                    current_state["message_count"],
-                    current_state["current_tool"],
-                    current_state["current_phase"],
-                    current_state["total_tool_calls"],
-                )
-                if state_key != last_write_state:
+                current_state_key = state_key(current_state)
+                if current_state_key != last_write_state:
                     atomic_write_json(current_state, progress_file)
-                    last_write_state = state_key
+                    last_write_state = current_state_key
             else:
                 idle_count += 1
-                # After 2 seconds of no new data, write current state anyway
-                # (ensures progress.json stays fresh)
-                if idle_count == 4:
+                # Every 2 seconds of no parent log data, refresh child Codex
+                # transcript stats and write if child activity advanced.
+                if idle_count % 4 == 0:
                     current_state = tracker.to_dict()
-                    atomic_write_json(current_state, progress_file)
+                    current_state_key = state_key(current_state)
+                    if current_state_key != last_write_state or idle_count == 4:
+                        atomic_write_json(current_state, progress_file)
+                        last_write_state = current_state_key
 
                 # After 3600 idle cycles (30 min), mark inactive and exit
                 if idle_count > 3600:

package/bundled/dev-pipeline-windows/.env.example

@@ -22,9 +22,10 @@
 # SESSION_TIMEOUT=0          # Session timeout in seconds (0 = no limit)
 # VERBOSE=1                  # Verbose logging (1=on, 0=off)
 # HEARTBEAT_INTERVAL=30      # Poll interval for session progress/stale checks
-# STALE_KILL_THRESHOLD=900   # Auto-kill session after N seconds without log progress (0 = disabled)
+# STALE_KILL_THRESHOLD=900   # Auto-kill session after N seconds without parent log progress (0 = disabled)
 # STALE_KILL_GRACE_SECONDS=10  # Grace period after stale-kill before force-stopping the job
-# CODEX_SUBAGENT_TIMEOUT_SECONDS=840  # Codex subagent max runtime; defaults to stale threshold - 60
+# CODEX_WAIT_STALE_KILL_THRESHOLD=3600  # Longer no-log window while Codex waits on subagents
+# CODEX_SUBAGENT_TIMEOUT_SECONDS=3300  # Codex subagent max runtime; defaults to wait threshold - 300
 # LOG_CLEANUP_ENABLED=1      # Run periodic session log cleanup
 # LOG_RETENTION_DAYS=14      # Delete session logs older than N days
 # LOG_MAX_TOTAL_MB=1024      # Keep total logs under N MB via oldest-first cleanup

package/bundled/dev-pipeline-windows/SCHEMA_ANALYSIS.md

@@ -346,9 +346,10 @@ pending, in_progress, completed, failed, skipped
 | `SESSION_TIMEOUT` | integer | 0 | 0 = no limit |
 | `VERBOSE` | integer | (not specified) | 1=on, 0=off |
 | `HEARTBEAT_INTERVAL` | integer | 30 | Poll interval for session progress/stale checks |
-| `STALE_KILL_THRESHOLD` | integer | 900 | Auto-kill after N seconds without log progress; 0 disables |
+| `STALE_KILL_THRESHOLD` | integer | 900 | Auto-kill after N seconds without parent log progress; 0 disables |
 | `STALE_KILL_GRACE_SECONDS` | integer | 10 | Grace period after stale-kill before force-stopping |
-| `CODEX_SUBAGENT_TIMEOUT_SECONDS` | integer | 840 | Codex subagent max runtime |
+| `CODEX_WAIT_STALE_KILL_THRESHOLD` | integer | 3600 | Longer no-log stale window while Codex waits on subagents |
+| `CODEX_SUBAGENT_TIMEOUT_SECONDS` | integer | 3300 | Codex subagent max runtime |
 | `LOG_CLEANUP_ENABLED` | boolean | 1 | Periodic session log cleanup |
 | `LOG_RETENTION_DAYS` | integer | 14 | Delete session logs older than N days |
 | `LOG_MAX_TOTAL_MB` | integer | 1024 | Keep total logs under N MB |

package/bundled/dev-pipeline-windows/lib/common.ps1

@@ -345,13 +345,86 @@ function Get-PrizmCodexSubagentTimeoutSeconds {
     return $configuredTimeout
   }
 
-  $outerThreshold = 0
-  $outerThresholdText = if ($env:STALE_KILL_THRESHOLD) { $env:STALE_KILL_THRESHOLD } else { $env:SESSION_TIMEOUT }
-  if ([int]::TryParse($outerThresholdText, [ref]$outerThreshold) -and $outerThreshold -gt 120) {
-    return ($outerThreshold - 60)
+  $waitThreshold = 0
+  $waitThresholdText = if ($env:CODEX_WAIT_STALE_KILL_THRESHOLD) { $env:CODEX_WAIT_STALE_KILL_THRESHOLD } else { '3600' }
+  if ([int]::TryParse($waitThresholdText, [ref]$waitThreshold) -and $waitThreshold -gt 600) {
+    return ($waitThreshold - 300)
   }
 
-  return 840
+  return 3300
+}
+
+function Get-PrizmEffectiveStaleKillThreshold {
+  param(
+    [string]$ProgressFile,
+    [int]$BaseThreshold
+  )
+
+  if ($BaseThreshold -le 0) { return $BaseThreshold }
+  if (-not (Test-Path $ProgressFile)) { return $BaseThreshold }
+
+  try {
+    $progress = Get-Content $ProgressFile -Raw -ErrorAction Stop | ConvertFrom-Json -ErrorAction Stop
+  } catch {
+    return $BaseThreshold
+  }
+
+  $spawnCount = 0
+  if ($progress.event_format -eq 'codex-json' -and $progress.current_tool -eq 'wait' -and $progress.tool_calls) {
+    foreach ($tool in @($progress.tool_calls)) {
+      if ($tool.name -eq 'spawn_agent') {
+        $count = 0
+        if ([int]::TryParse([string]$tool.count, [ref]$count)) { $spawnCount += $count }
+      }
+    }
+  }
+
+  if ($spawnCount -le 0) { return $BaseThreshold }
+
+  $waitThreshold = 0
+  if ([int]::TryParse($env:CODEX_WAIT_STALE_KILL_THRESHOLD, [ref]$waitThreshold) -and $waitThreshold -gt $BaseThreshold) {
+    return $waitThreshold
+  }
+
+  return [Math]::Max($BaseThreshold * 4, 3600)
+}
+
+function Get-PrizmProgressChildActivity {
+  param([string]$ProgressFile)
+
+  $empty = [pscustomobject]@{
+    Signature = ''
+    TotalBytes = 0
+    SessionCount = 0
+  }
+  if (-not (Test-Path $ProgressFile)) { return $empty }
+
+  try {
+    $progress = Get-Content $ProgressFile -Raw -ErrorAction Stop | ConvertFrom-Json -ErrorAction Stop
+  } catch {
+    return $empty
+  }
+
+  $signature = ''
+  if ($progress.PSObject.Properties['child_activity_signature'] -and $progress.child_activity_signature) {
+    $signature = [string]$progress.child_activity_signature
+  }
+
+  $totalBytes = [int64]0
+  if ($progress.PSObject.Properties['child_total_bytes']) {
+    [int64]::TryParse([string]$progress.child_total_bytes, [ref]$totalBytes) | Out-Null
+  }
+
+  $sessionCount = 0
+  if ($progress.PSObject.Properties['child_session_files'] -and $progress.child_session_files) {
+    $sessionCount = @($progress.child_session_files).Count
+  }
+
+  return [pscustomobject]@{
+    Signature = $signature
+    TotalBytes = $totalBytes
+    SessionCount = $sessionCount
+  }
 }
 
 function Test-PrizmCodexJsonSupport {

package/bundled/dev-pipeline-windows/lib/pipeline.ps1

@@ -552,6 +552,7 @@ function Invoke-PrizmPipeline {
     $elapsedSeconds = 0
     $staleSeconds = 0
     $previousLogSize = 0
+    $previousChildActivitySignature = ''
     $wasTimedOut = $false
     $staleKillMarker = Join-Path $logsDir 'stale-kill.json'
     $wasStaleKilled = $false
@@ -568,7 +569,13 @@ function Invoke-PrizmPipeline {
       }
       $growth = $currentLogSize - $previousLogSize
       $previousLogSize = $currentLogSize
-      if ($growth -gt 0) {
+
+      $childActivity = Get-PrizmProgressChildActivity -ProgressFile $progressJson
+      $childSignature = [string]$childActivity.Signature
+      $childAdvanced = ($childSignature -and $childSignature -ne $previousChildActivitySignature)
+      $previousChildActivitySignature = $childSignature
+
+      if ($growth -gt 0 -or $childAdvanced) {
         $staleSeconds = 0
       } else {
         $staleSeconds += $waitSeconds
@@ -580,10 +587,11 @@ function Invoke-PrizmPipeline {
         break
       }
 
-      if ($staleKillThreshold -gt 0 -and $staleSeconds -ge $staleKillThreshold) {
+      $effectiveStaleKillThreshold = Get-PrizmEffectiveStaleKillThreshold -ProgressFile $progressJson -BaseThreshold $staleKillThreshold
+      if ($effectiveStaleKillThreshold -gt 0 -and $staleSeconds -ge $effectiveStaleKillThreshold) {
         $wasStaleKilled = $true
-        Write-PrizmWarn "Session stale-killed (no progress for ${staleKillThreshold}s)"
-        Write-PrizmStaleKillMarker $staleKillMarker $staleSeconds $staleKillThreshold
+        Write-PrizmWarn "Session stale-killed (no progress for ${effectiveStaleKillThreshold}s)"
+        Write-PrizmStaleKillMarker $staleKillMarker $staleSeconds $effectiveStaleKillThreshold
         Stop-PrizmSessionProcess $pidPath
         if ($staleKillGraceSeconds -gt 0) { Start-Sleep -Seconds $staleKillGraceSeconds }
         break

package/bundled/dev-pipeline-windows/run-recovery.ps1

@@ -110,6 +110,7 @@ $job = Start-Job -ScriptBlock {
 $elapsedSeconds = 0
 $staleSeconds = 0
 $previousLogSize = 0
+$previousChildActivitySignature = ''
 $wasTimedOut = $false
 $wasStaleKilled = $false
 while ($true) {
@@ -123,7 +124,13 @@ while ($true) {
   if (Test-Path $logPath) { $currentLogSize = [int64](Get-Item $logPath).Length }
   $growth = $currentLogSize - $previousLogSize
   $previousLogSize = $currentLogSize
-  if ($growth -gt 0) { $staleSeconds = 0 } else { $staleSeconds += $waitSeconds }
+
+  $childActivity = Get-PrizmProgressChildActivity -ProgressFile $progressPath
+  $childSignature = [string]$childActivity.Signature
+  $childAdvanced = ($childSignature -and $childSignature -ne $previousChildActivitySignature)
+  $previousChildActivitySignature = $childSignature
+
+  if ($growth -gt 0 -or $childAdvanced) { $staleSeconds = 0 } else { $staleSeconds += $waitSeconds }
 
   if ($timeoutSeconds -gt 0 -and $elapsedSeconds -ge $timeoutSeconds) {
     $wasTimedOut = $true

package/bundled/dev-pipeline-windows/scripts/parse-stream-progress.py

@@ -23,6 +23,7 @@ import tempfile
 import time
 from collections import Counter
 from datetime import datetime, timezone
+from pathlib import Path
 
 
 # Ordered pipeline phases — index defines forward-only progression.
@@ -76,6 +77,13 @@ class ProgressTracker:
         self.event_format = ""
         self.active_subagent_count = 0
         self.subagent_status_counts = Counter()
+        self.codex_child_thread_ids = set()
+        self.child_session_files = []
+        self.child_total_bytes = 0
+        self.child_activity_signature = ""
+        self.last_child_activity_at = ""
+        self._codex_child_session_paths = {}
+        self._last_child_scan_at = 0.0
         self._text_buffer = ""
         self._in_tool_use = False
         self._current_tool_input_parts = []
@@ -113,6 +121,9 @@ class ProgressTracker:
 
                 elif item_type == "collab_tool_call":
                     tool_name = item.get("tool", "collab")
+                    self._record_codex_child_thread_ids(
+                        item.get("receiver_thread_ids")
+                    )
                     if event_type == "item.started":
                         self.current_tool = tool_name
                         self.tool_call_counts[tool_name] += 1
@@ -345,8 +356,117 @@ class ProgressTracker:
         self.subagent_status_counts = counts
         self.active_subagent_count = active
 
+    def _record_codex_child_thread_ids(self, thread_ids):
+        """Remember Codex child thread IDs reported by collab tool calls."""
+        if not isinstance(thread_ids, list):
+            return
+        for thread_id in thread_ids:
+            if isinstance(thread_id, str) and thread_id.strip():
+                self.codex_child_thread_ids.add(thread_id.strip())
+
+    def _codex_sessions_dir(self):
+        """Return the Codex sessions directory for the current environment."""
+        codex_home = os.environ.get("CODEX_HOME")
+        if codex_home:
+            return Path(codex_home).expanduser() / "sessions"
+        return Path.home() / ".codex" / "sessions"
+
+    def _find_codex_child_session_file(self, thread_id):
+        """Find a Codex transcript file for a child thread ID."""
+        sessions_dir = self._codex_sessions_dir()
+        if not sessions_dir.exists():
+            return None
+
+        try:
+            matches = list(sessions_dir.rglob(f"*{thread_id}.jsonl"))
+        except OSError:
+            return None
+
+        if not matches:
+            return None
+
+        try:
+            matches.sort(key=lambda path: path.stat().st_mtime, reverse=True)
+        except OSError:
+            pass
+        return str(matches[0])
+
+    def refresh_child_session_activity(self, force=False):
+        """Refresh Codex child transcript file stats.
+
+        The heartbeat monitor uses this activity signature to treat subagent
+        transcript growth as real progress while the parent Codex session is
+        blocked in `wait`.
+        """
+        previous_signature = self.child_activity_signature
+
+        if not self.codex_child_thread_ids:
+            self.child_session_files = []
+            self.child_total_bytes = 0
+            self.child_activity_signature = ""
+            self.last_child_activity_at = ""
+            return previous_signature != self.child_activity_signature
+
+        now = time.monotonic()
+        should_scan = (
+            force
+            or self._last_child_scan_at == 0.0
+            or (now - self._last_child_scan_at >= 2.0)
+        )
+        if should_scan:
+            for thread_id in sorted(self.codex_child_thread_ids):
+                path = self._codex_child_session_paths.get(thread_id)
+                if not path or not os.path.exists(path):
+                    found = self._find_codex_child_session_file(thread_id)
+                    if found:
+                        self._codex_child_session_paths[thread_id] = found
+            self._last_child_scan_at = now
+
+        files = []
+        signature_parts = []
+        total_bytes = 0
+        max_mtime = 0.0
+
+        for thread_id in sorted(self.codex_child_thread_ids):
+            path = self._codex_child_session_paths.get(thread_id)
+            if not path:
+                continue
+            try:
+                stat = os.stat(path)
+            except OSError:
+                continue
+
+            total_bytes += stat.st_size
+            max_mtime = max(max_mtime, stat.st_mtime)
+            signature_parts.append(
+                f"{thread_id}:{stat.st_size}:{getattr(stat, 'st_mtime_ns', int(stat.st_mtime * 1_000_000_000))}"
+            )
+            files.append(
+                {
+                    "thread_id": thread_id,
+                    "path": path,
+                    "size": stat.st_size,
+                    "mtime": datetime.fromtimestamp(
+                        stat.st_mtime, timezone.utc
+                    ).strftime("%Y-%m-%dT%H:%M:%SZ"),
+                }
+            )
+
+        self.child_session_files = files
+        self.child_total_bytes = total_bytes
+        self.child_activity_signature = "|".join(signature_parts)
+        self.last_child_activity_at = (
+            datetime.fromtimestamp(max_mtime, timezone.utc).strftime(
+                "%Y-%m-%dT%H:%M:%SZ"
+            )
+            if max_mtime
+            else ""
+        )
+        return previous_signature != self.child_activity_signature
+
     def to_dict(self):
         """Export current state as a dictionary for JSON serialization."""
+        self.refresh_child_session_activity()
         tool_calls = [
             {"name": name, "count": count}
             for name, count in self.tool_call_counts.most_common()
@@ -367,6 +487,11 @@ class ProgressTracker:
             "total_tool_calls": self.total_tool_calls,
             "active_subagent_count": self.active_subagent_count,
             "subagent_states": subagent_states,
+            "child_thread_ids": sorted(self.codex_child_thread_ids),
+            "child_session_files": self.child_session_files,
+            "child_total_bytes": self.child_total_bytes,
+            "child_activity_signature": self.child_activity_signature,
+            "last_child_activity_at": self.last_child_activity_at,
             "last_text_snippet": self.last_text_snippet,
             "is_active": self.is_active,
             "errors": self.errors[-10:],  # Keep last 10 errors
@@ -397,6 +522,15 @@ def tail_and_parse(session_log, progress_file, poll_interval=0.5):
     tracker = ProgressTracker()
     last_write_state = None
 
+    def state_key(state):
+        return (
+            state["message_count"],
+            state["current_tool"],
+            state["current_phase"],
+            state["total_tool_calls"],
+            state.get("child_activity_signature", ""),
+        )
+
     # Wait for log file to appear
     wait_count = 0
     while not os.path.exists(session_log):
@@ -428,22 +562,20 @@ def tail_and_parse(session_log, progress_file, poll_interval=0.5):
 
                 # Write progress if state changed
                 current_state = tracker.to_dict()
-                state_key = (
-                    current_state["message_count"],
-                    current_state["current_tool"],
-                    current_state["current_phase"],
-                    current_state["total_tool_calls"],
-                )
-                if state_key != last_write_state:
+                current_state_key = state_key(current_state)
+                if current_state_key != last_write_state:
                     atomic_write_json(current_state, progress_file)
-                    last_write_state = state_key
+                    last_write_state = current_state_key
             else:
                 idle_count += 1
-                # After 2 seconds of no new data, write current state anyway
-                # (ensures progress.json stays fresh)
-                if idle_count == 4:
+                # Every 2 seconds of no parent log data, refresh child Codex
+                # transcript stats and write if child activity advanced.
+                if idle_count % 4 == 0:
                     current_state = tracker.to_dict()
-                    atomic_write_json(current_state, progress_file)
+                    current_state_key = state_key(current_state)
+                    if current_state_key != last_write_state or idle_count == 4:
+                        atomic_write_json(current_state, progress_file)
+                        last_write_state = current_state_key
 
                 # After 3600 idle cycles (30 min), mark inactive and exit
                 if idle_count > 3600:

package/bundled/skills/_metadata.json

@@ -1,5 +1,5 @@
 {
-  "version": "1.1.66",
+  "version": "1.1.67",
   "skills": {
     "prizm-kit": {
       "description": "Full-lifecycle dev toolkit. Covers spec-driven development, Prizm context docs, code quality, debugging, deployment, and knowledge management.",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "prizmkit",
-  "version": "1.1.66",
+  "version": "1.1.67",
   "description": "Create a new PrizmKit-powered project with clean initialization — no framework dev files, just what you need.",
   "type": "module",
   "bin": {

package/src/scaffold.js

@@ -576,7 +576,7 @@ project_doc_fallback_filenames = ["CLAUDE.md", "CODEBUDDY.md"]
 
 [agents]
 max_depth = 1
-job_max_runtime_seconds = 840
+job_max_runtime_seconds = 3300
 `;
     await fs.writeFile(configPath, configToml);
     await fs.remove(legacySettingsPath);