prizmkit 1.1.21 → 1.1.24

package/bundled/VERSION.json

@@ -1,5 +1,5 @@
 {
-  "frameworkVersion": "1.1.21",
-  "bundledAt": "2026-04-11T11:23:13.654Z",
-  "bundledFrom": "5137496"
+  "frameworkVersion": "1.1.24",
+  "bundledAt": "2026-04-12T01:13:39.954Z",
+  "bundledFrom": "626fbf5"
 }

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

@@ -6,9 +6,14 @@
 # structured progress from progress.json (written by
 # parse-stream-progress.py) and fall back to tail-based monitoring.
 #
+# When stale_kill_threshold is set (>0), the heartbeat monitor will
+# automatically kill the AI CLI process if it shows no progress for
+# the specified duration. This prevents sessions from hanging forever
+# when the AI CLI process doesn't exit after completing its work.
+#
 # Usage:
 #   source "$SCRIPT_DIR/lib/heartbeat.sh"
-#   start_heartbeat "$cli_pid" "$session_log" "$progress_json" "$interval"
+#   start_heartbeat "$cli_pid" "$session_log" "$progress_json" "$interval" ["$stale_kill_threshold"]
 #   # ... wait for CLI to finish ...
 #   stop_heartbeat "$_HEARTBEAT_PID"
 #
@@ -20,19 +25,23 @@
 # Sets _HEARTBEAT_PID to the background process PID.
 #
 # Arguments:
-#   $1 - cli_pid         PID of the AI CLI process to monitor
-#   $2 - session_log     Path to session.log
-#   $3 - progress_json   Path to progress.json (may not exist if stream-json disabled)
-#   $4 - interval        Heartbeat interval in seconds
+#   $1 - cli_pid              PID of the AI CLI process to monitor
+#   $2 - session_log          Path to session.log
+#   $3 - progress_json        Path to progress.json (may not exist if stream-json disabled)
+#   $4 - interval             Heartbeat interval in seconds
+#   $5 - stale_kill_threshold (optional) Seconds of no progress before auto-killing the process.
+#                              0 = disabled (default). Recommended: 900 (15 minutes).
 start_heartbeat() {
     local cli_pid="$1"
     local session_log="$2"
     local progress_json="$3"
     local heartbeat_interval="$4"
+    local stale_kill_threshold="${5:-0}"
 
     (
         local elapsed=0
         local prev_size=0
+        local stale_seconds=0
         while kill -0 "$cli_pid" 2>/dev/null; do
             sleep "$heartbeat_interval"
             elapsed=$((elapsed + heartbeat_interval))
@@ -48,6 +57,13 @@ start_heartbeat() {
             local growth=$((cur_size - prev_size))
             prev_size=$cur_size
 
+            # Track progress staleness (no log growth = stale)
+            if [[ $growth -eq 0 ]]; then
+                stale_seconds=$((stale_seconds + heartbeat_interval))
+            else
+                stale_seconds=0
+            fi
+
             local size_display
             if [[ $cur_size -gt 1048576 ]]; then
                 size_display="$((cur_size / 1048576))MB"
@@ -67,6 +83,33 @@ start_heartbeat() {
                 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 $stale_kill_threshold ]]; then
+                local stale_mins=$((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} Killing AI CLI process $cli_pid (stale session)..."
+                kill -TERM "$cli_pid" 2>/dev/null || true
+                # Give process 10s to exit gracefully, then force kill
+                sleep 10
+                if kill -0 "$cli_pid" 2>/dev/null; then
+                    echo -e "  ${RED}[HEARTBEAT]${NC} Process still alive after SIGTERM, sending SIGKILL..."
+                    kill -9 "$cli_pid" 2>/dev/null || true
+                fi
+                # Write stale-kill marker so spawn_and_wait_session knows this wasn't a crash
+                local _marker_dir
+                _marker_dir="$(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
+                break
+            fi
+
+            # Build staleness hint for display
+            local stale_hint=""
+            if [[ $stale_kill_threshold -gt 0 && $stale_seconds -gt 0 ]]; then
+                local stale_mins=$((stale_seconds / 60))
+                local threshold_mins=$((stale_kill_threshold / 60))
+                stale_hint=" | stale: ${stale_mins}m/${threshold_mins}m"
+            fi
+
             # Try structured progress from progress.json
             if [[ -f "$progress_json" ]]; then
                 local phase tool msgs tools_total
@@ -86,7 +129,7 @@ try:
 except Exception:
     sys.exit(1)
 " "$progress_json" 2>/dev/null) && {
-                    echo -e "  ${status_icon} ${BLUE}[HEARTBEAT]${NC} ${mins}m${secs}s | log: ${size_display} | ${phase}"
+                    echo -e "  ${status_icon} ${BLUE}[HEARTBEAT]${NC} ${mins}m${secs}s | log: ${size_display} | ${phase}${stale_hint}"
                     continue
                 }
             fi
@@ -97,7 +140,7 @@ 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}"
+            echo -e "  ${status_icon} ${BLUE}[HEARTBEAT]${NC} ${mins}m${secs}s elapsed | log: ${size_display} (+${growth}B) | ${last_activity}${stale_hint}"
         done
     ) &
     _HEARTBEAT_PID=$!

package/bundled/dev-pipeline/reset-bug.sh

@@ -139,22 +139,20 @@ fi
 BUG_IDS=()
 
 if [[ -n "$FILTER_MODE" ]]; then
-    # Filter by status from .prizmkit/state/bugfix/bugs/*/status.json
+    # Filter by status from bug-fix-list.json (single source of truth)
     while IFS= read -r bid; do
         [[ -n "$bid" ]] && BUG_IDS+=("$bid")
     done < <(python3 -c "
-import json, os, sys
-state_dir = '$STATE_DIR'
+import json, sys
 filter_mode = '$FILTER_MODE'
-bugs_dir = os.path.join(state_dir, 'bugs')
-if not os.path.isdir(bugs_dir):
-    sys.exit(0)
-for bid in sorted(os.listdir(bugs_dir)):
-    status_file = os.path.join(bugs_dir, bid, 'status.json')
-    if not os.path.isfile(status_file):
+bug_list = '$BUG_LIST'
+with open(bug_list) as f:
+    data = json.load(f)
+for bug in data.get('bugs', []):
+    if not isinstance(bug, dict):
         continue
-    with open(status_file) as f:
-        status = json.load(f).get('status', '')
+    bid = bug.get('id', '')
+    status = bug.get('status', '')
     if filter_mode == 'auto_skipped' and status == 'auto_skipped':
         print(bid)
     elif filter_mode == 'failed' and status == 'failed':
@@ -244,13 +242,23 @@ sys.exit(1)
     echo -e "${BOLD}════════════════════════════════════════════════════${NC}"
 
     STATUS_FILE="$STATE_DIR/bugs/$CUR_BUG_ID/status.json"
+    # Read status from bug-fix-list.json (single source of truth)
+    CURRENT_STATUS=$(python3 -c "
+import json, sys
+with open('$BUG_LIST') as f:
+    data = json.load(f)
+for bug in data.get('bugs', []):
+    if isinstance(bug, dict) and bug.get('id') == '$CUR_BUG_ID':
+        print(bug.get('status', '?'))
+        sys.exit(0)
+print('?')
+" 2>/dev/null || echo "?")
     if [[ -f "$STATUS_FILE" ]]; then
-        CURRENT_STATUS=$(python3 -c "import json; d=json.load(open('$STATUS_FILE')); print(d.get('status','?'))")
         CURRENT_RETRY=$(python3 -c "import json; d=json.load(open('$STATUS_FILE')); print(d.get('retry_count',0))")
         SESSION_COUNT=$(python3 -c "import json; d=json.load(open('$STATUS_FILE')); print(len(d.get('sessions',[])))")
         log_info "Current status: $CURRENT_STATUS (retry $CURRENT_RETRY, $SESSION_COUNT sessions)"
     else
-        log_info "No status file found (never executed)"
+        log_info "Current status: $CURRENT_STATUS (no runtime state file)"
     fi
 
     BUGFIX_DIR="$PROJECT_ROOT/.prizmkit/bugfix/$CUR_BUG_ID"

package/bundled/dev-pipeline/reset-feature.sh

@@ -139,22 +139,20 @@ fi
 FEATURE_IDS=()
 
 if [[ -n "$FILTER_MODE" ]]; then
-    # Filter by status from state/features/*/status.json
+    # Filter by status from feature-list.json (single source of truth)
     while IFS= read -r fid; do
         [[ -n "$fid" ]] && FEATURE_IDS+=("$fid")
     done < <(python3 -c "
-import json, os, sys
-state_dir = '$STATE_DIR'
+import json, sys
 filter_mode = '$FILTER_MODE'
-features_dir = os.path.join(state_dir, 'features')
-if not os.path.isdir(features_dir):
-    sys.exit(0)
-for fid in sorted(os.listdir(features_dir)):
-    status_file = os.path.join(features_dir, fid, 'status.json')
-    if not os.path.isfile(status_file):
+feature_list = '$FEATURE_LIST'
+with open(feature_list) as f:
+    data = json.load(f)
+for feat in data.get('features', []):
+    if not isinstance(feat, dict):
         continue
-    with open(status_file) as f:
-        status = json.load(f).get('status', '')
+    fid = feat.get('id', '')
+    status = feat.get('status', '')
     if filter_mode == 'auto_skipped' and status == 'auto_skipped':
         print(fid)
     elif filter_mode == 'failed' and status == 'failed':
@@ -253,13 +251,23 @@ sys.exit(1)
     echo -e "${BOLD}════════════════════════════════════════════════════${NC}"
 
     STATUS_FILE="$STATE_DIR/features/$CUR_FEATURE_ID/status.json"
+    # Read status from feature-list.json (single source of truth)
+    CURRENT_STATUS=$(python3 -c "
+import json, sys
+with open('$FEATURE_LIST') as f:
+    data = json.load(f)
+for feat in data.get('features', []):
+    if isinstance(feat, dict) and feat.get('id') == '$CUR_FEATURE_ID':
+        print(feat.get('status', '?'))
+        sys.exit(0)
+print('?')
+" 2>/dev/null || echo "?")
     if [[ -f "$STATUS_FILE" ]]; then
-        CURRENT_STATUS=$(python3 -c "import json; d=json.load(open('$STATUS_FILE')); print(d.get('status','?'))")
         CURRENT_RETRY=$(python3 -c "import json; d=json.load(open('$STATUS_FILE')); print(d.get('retry_count',0))")
         SESSION_COUNT=$(python3 -c "import json; d=json.load(open('$STATUS_FILE')); print(len(d.get('sessions',[])))")
         log_info "Current status: $CURRENT_STATUS (retry $CURRENT_RETRY, $SESSION_COUNT sessions)"
     else
-        log_info "No status file found (never executed)"
+        log_info "Current status: $CURRENT_STATUS (no runtime state file)"
     fi
 
     SPECS_DIR="$PROJECT_ROOT/.prizmkit/specs/$FEATURE_SLUG"

package/bundled/dev-pipeline/reset-refactor.sh

@@ -129,22 +129,20 @@ fi
 REFACTOR_IDS=()
 
 if [[ -n "$FILTER_MODE" ]]; then
-    # Filter by status from .prizmkit/state/refactor/refactors/*/status.json
+    # Filter by status from refactor-list.json (single source of truth)
     while IFS= read -r rid; do
         [[ -n "$rid" ]] && REFACTOR_IDS+=("$rid")
     done < <(python3 -c "
-import json, os, sys
-state_dir = '$STATE_DIR'
+import json, sys
 filter_mode = '$FILTER_MODE'
-refactors_dir = os.path.join(state_dir, 'refactors')
-if not os.path.isdir(refactors_dir):
-    sys.exit(0)
-for rid in sorted(os.listdir(refactors_dir)):
-    status_file = os.path.join(refactors_dir, rid, 'status.json')
-    if not os.path.isfile(status_file):
+refactor_list = '$REFACTOR_LIST'
+with open(refactor_list) as f:
+    data = json.load(f)
+for r in data.get('refactors', []):
+    if not isinstance(r, dict):
         continue
-    with open(status_file) as f:
-        status = json.load(f).get('status', '')
+    rid = r.get('id', '')
+    status = r.get('status', '')
     if filter_mode == 'auto_skipped' and status == 'auto_skipped':
         print(rid)
     elif filter_mode == 'failed' and status == 'failed':
@@ -242,13 +240,23 @@ sys.exit(1)
     echo -e "${BOLD}════════════════════════════════════════════════════${NC}"
 
     STATUS_FILE="$STATE_DIR/refactors/$CUR_REFACTOR_ID/status.json"
+    # Read status from refactor-list.json (single source of truth)
+    CURRENT_STATUS=$(python3 -c "
+import json, sys
+with open('$REFACTOR_LIST') as f:
+    data = json.load(f)
+for r in data.get('refactors', []):
+    if isinstance(r, dict) and r.get('id') == '$CUR_REFACTOR_ID':
+        print(r.get('status', '?'))
+        sys.exit(0)
+print('?')
+" 2>/dev/null || echo "?")
     if [[ -f "$STATUS_FILE" ]]; then
-        CURRENT_STATUS=$(python3 -c "import json; d=json.load(open('$STATUS_FILE')); print(d.get('status','?'))")
         CURRENT_RETRY=$(python3 -c "import json; d=json.load(open('$STATUS_FILE')); print(d.get('retry_count',0))")
         SESSION_COUNT=$(python3 -c "import json; d=json.load(open('$STATUS_FILE')); print(len(d.get('sessions',[])))")
         log_info "Current status: $CURRENT_STATUS (retry $CURRENT_RETRY, $SESSION_COUNT sessions)"
     else
-        log_info "No status file found (never executed)"
+        log_info "Current status: $CURRENT_STATUS (no runtime state file)"
     fi
 
     SPECS_DIR="$PROJECT_ROOT/.prizmkit/specs/$REFACTOR_SLUG"

package/bundled/dev-pipeline/run-bugfix.sh

@@ -21,6 +21,7 @@ set -euo pipefail
 #   PRIZMKIT_PLATFORM     Force platform: 'codebuddy' or 'claude' (auto-detected)
 #   VERBOSE               Set to 1 to enable --verbose on AI CLI
 #   HEARTBEAT_INTERVAL    Heartbeat log interval in seconds (default: 30)
+#   STALE_KILL_THRESHOLD   Auto-kill session after N seconds of no progress (default: 900)
 #   HEARTBEAT_STALE_THRESHOLD  Heartbeat stale threshold in seconds (default: 600)
 #   LOG_CLEANUP_ENABLED   Run periodic log cleanup (default: 1)
 #   LOG_RETENTION_DAYS    Delete logs older than N days (default: 14)
@@ -39,6 +40,7 @@ MAX_RETRIES=${MAX_RETRIES:-3}
 SESSION_TIMEOUT=${SESSION_TIMEOUT:-0}
 HEARTBEAT_STALE_THRESHOLD=${HEARTBEAT_STALE_THRESHOLD:-600}
 HEARTBEAT_INTERVAL=${HEARTBEAT_INTERVAL:-30}
+STALE_KILL_THRESHOLD=${STALE_KILL_THRESHOLD:-900}
 LOG_CLEANUP_ENABLED=${LOG_CLEANUP_ENABLED:-1}
 LOG_RETENTION_DAYS=${LOG_RETENTION_DAYS:-14}
 LOG_MAX_TOTAL_MB=${LOG_MAX_TOTAL_MB:-1024}
@@ -146,8 +148,8 @@ spawn_and_wait_session() {
         watcher_pid=$!
     fi
 
-    # Heartbeat monitor
-    start_heartbeat "$cli_pid" "$session_log" "$progress_json" "$HEARTBEAT_INTERVAL"
+    # Heartbeat monitor (with stale-kill protection)
+    start_heartbeat "$cli_pid" "$session_log" "$progress_json" "$HEARTBEAT_INTERVAL" "$STALE_KILL_THRESHOLD"
     local heartbeat_pid="${_HEARTBEAT_PID:-}"
 
     # Wait for AI CLI to finish
@@ -166,6 +168,14 @@ spawn_and_wait_session() {
 
     [[ $exit_code -eq 143 ]] && exit_code=124
 
+    # Check for stale-kill marker (heartbeat killed the process due to no progress)
+    local stale_kill_marker="$session_dir/logs/stale-kill.json"
+    local was_stale_killed=false
+    if [[ -f "$stale_kill_marker" ]]; then
+        was_stale_killed=true
+        log_warn "Session was stale-killed by heartbeat monitor (no progress for too long)"
+    fi
+
     # Session summary
     if [[ -f "$session_log" ]]; then
         local final_size=$(wc -c < "$session_log" 2>/dev/null | tr -d ' ')
@@ -183,6 +193,33 @@ spawn_and_wait_session() {
     if [[ $exit_code -eq 124 ]]; then
         log_warn "Session timed out after ${SESSION_TIMEOUT}s"
         session_status="timed_out"
+    elif [[ "$was_stale_killed" == true ]]; then
+        log_warn "Session stale-killed (no progress for ${STALE_KILL_THRESHOLD}s)"
+        local has_commits=""
+        if git -C "$project_root" rev-parse --is-inside-work-tree >/dev/null 2>&1; then
+            has_commits=$(git -C "$project_root" log "${default_branch}..HEAD" --oneline 2>/dev/null | head -1)
+        fi
+        if [[ -n "$has_commits" ]]; then
+            log_info "Stale-killed session has commits — treating as success"
+            session_status="success"
+        else
+            local uncommitted=""
+            uncommitted=$(git -C "$project_root" status --porcelain 2>/dev/null | head -1 || true)
+            if [[ -n "$uncommitted" ]]; then
+                log_warn "Stale-killed session has uncommitted changes — auto-committing..."
+                git -C "$project_root" add -A 2>/dev/null || true
+                if git -C "$project_root" commit --no-verify -m "chore($bug_id): auto-commit session work (stale-killed)" 2>/dev/null; then
+                    log_info "Auto-commit succeeded"
+                    session_status="success"
+                else
+                    log_warn "Auto-commit failed — no changes to commit"
+                    session_status="crashed"
+                fi
+            else
+                log_warn "Stale-killed session produced no commits and no changes"
+                session_status="crashed"
+            fi
+        fi
     elif [[ $exit_code -ne 0 ]]; then
         log_warn "Session exited with code $exit_code"
         session_status="crashed"
@@ -1091,6 +1128,7 @@ show_help() {
     echo "  AI_CLI                AI CLI command name (auto-detected: cbc or claude)"
     echo "  VERBOSE               Set to 1 for verbose AI CLI output"
     echo "  HEARTBEAT_INTERVAL    Heartbeat log interval in seconds (default: 30)"
+    echo "  STALE_KILL_THRESHOLD  Auto-kill session after N seconds of no progress (default: 900)"
     echo "  LOG_CLEANUP_ENABLED   Run log cleanup before execution (default: 1)"
     echo "  LOG_RETENTION_DAYS    Delete logs older than N days (default: 14)"
     echo "  LOG_MAX_TOTAL_MB      Keep total logs under N MB (default: 1024)"

package/bundled/dev-pipeline/run-feature.sh

@@ -23,6 +23,7 @@ set -euo pipefail
 #   MODEL                 AI model to use (e.g. claude-opus-4.6, claude-sonnet-4.6, claude-haiku-4.5)
 #   VERBOSE               Set to 1 to enable --verbose on AI CLI (shows subagent output)
 #   HEARTBEAT_INTERVAL    Heartbeat log interval in seconds (default: 30)
+#   STALE_KILL_THRESHOLD   Auto-kill session after N seconds of no progress (default: 900)
 #   HEARTBEAT_STALE_THRESHOLD  Heartbeat stale threshold in seconds (default: 600)
 #   LOG_CLEANUP_ENABLED   Run periodic log cleanup (default: 1)
 #   LOG_RETENTION_DAYS    Delete logs older than N days (default: 14)
@@ -42,6 +43,7 @@ MAX_RETRIES=${MAX_RETRIES:-3}
 SESSION_TIMEOUT=${SESSION_TIMEOUT:-0}
 HEARTBEAT_STALE_THRESHOLD=${HEARTBEAT_STALE_THRESHOLD:-600}
 HEARTBEAT_INTERVAL=${HEARTBEAT_INTERVAL:-30}
+STALE_KILL_THRESHOLD=${STALE_KILL_THRESHOLD:-900}
 LOG_CLEANUP_ENABLED=${LOG_CLEANUP_ENABLED:-1}
 LOG_RETENTION_DAYS=${LOG_RETENTION_DAYS:-14}
 LOG_MAX_TOTAL_MB=${LOG_MAX_TOTAL_MB:-1024}
@@ -157,7 +159,8 @@ spawn_and_wait_session() {
     fi
 
     # Heartbeat monitor (reads progress.json when available, falls back to tail)
-    start_heartbeat "$cbc_pid" "$session_log" "$progress_json" "$HEARTBEAT_INTERVAL"
+    # Also monitors for stale sessions and auto-kills if no progress for STALE_KILL_THRESHOLD seconds
+    start_heartbeat "$cbc_pid" "$session_log" "$progress_json" "$HEARTBEAT_INTERVAL" "$STALE_KILL_THRESHOLD"
     local heartbeat_pid="${_HEARTBEAT_PID:-}"
 
     # Wait for AI CLI to finish
@@ -179,6 +182,14 @@ spawn_and_wait_session() {
         exit_code=124
     fi
 
+    # Check for stale-kill marker (heartbeat killed the process due to no progress)
+    local stale_kill_marker="$session_dir/logs/stale-kill.json"
+    local was_stale_killed=false
+    if [[ -f "$stale_kill_marker" ]]; then
+        was_stale_killed=true
+        log_warn "Session was stale-killed by heartbeat monitor (no progress for too long)"
+    fi
+
     # Show final session summary
     if [[ -f "$session_log" ]]; then
         local final_size=$(wc -c < "$session_log" 2>/dev/null | tr -d ' ')
@@ -198,6 +209,34 @@ spawn_and_wait_session() {
     if [[ $exit_code -eq 124 ]]; then
         log_warn "Session timed out after ${SESSION_TIMEOUT}s"
         session_status="timed_out"
+    elif [[ "$was_stale_killed" == true ]]; then
+        log_warn "Session stale-killed (no progress for ${STALE_KILL_THRESHOLD}s)"
+        # Treat stale-killed as potentially successful — check for commits
+        local has_commits=""
+        if git -C "$project_root" rev-parse --is-inside-work-tree >/dev/null 2>&1; then
+            has_commits=$(git -C "$project_root" log "${default_branch}..HEAD" --oneline 2>/dev/null | head -1)
+        fi
+        if [[ -n "$has_commits" ]]; then
+            log_info "Stale-killed session has commits — treating as success"
+            session_status="success"
+        else
+            local uncommitted=""
+            uncommitted=$(git -C "$project_root" status --porcelain 2>/dev/null | head -1 || true)
+            if [[ -n "$uncommitted" ]]; then
+                log_warn "Stale-killed session has uncommitted changes — auto-committing..."
+                git -C "$project_root" add -A 2>/dev/null || true
+                if git -C "$project_root" commit --no-verify -m "chore($feature_id): auto-commit session work (stale-killed)" 2>/dev/null; then
+                    log_info "Auto-commit succeeded"
+                    session_status="success"
+                else
+                    log_warn "Auto-commit failed — no changes to commit"
+                    session_status="crashed"
+                fi
+            else
+                log_warn "Stale-killed session produced no commits and no changes"
+                session_status="crashed"
+            fi
+        fi
     elif [[ $exit_code -ne 0 ]]; then
         log_warn "Session exited with code $exit_code"
         session_status="crashed"
@@ -1318,6 +1357,7 @@ show_help() {
     echo "  AI_CLI                AI CLI command name (auto-detected: cbc or claude)"
     echo "  MODEL                 AI model ID (e.g. claude-opus-4.6, claude-sonnet-4.6, claude-haiku-4.5)"
     echo "  HEARTBEAT_INTERVAL    Heartbeat log interval in seconds (default: 30)"
+    echo "  STALE_KILL_THRESHOLD  Auto-kill session after N seconds of no progress (default: 900)"
     echo "  HEARTBEAT_STALE_THRESHOLD  Heartbeat stale threshold in seconds (default: 600)"
     echo "  LOG_CLEANUP_ENABLED   Run log cleanup before execution (default: 1)"
     echo "  LOG_RETENTION_DAYS    Delete logs older than N days (default: 14)"

package/bundled/dev-pipeline/run-refactor.sh

@@ -21,6 +21,7 @@ set -euo pipefail
 #   PRIZMKIT_PLATFORM     Force platform: 'codebuddy' or 'claude' (auto-detected)
 #   VERBOSE               Set to 1 to enable --verbose on AI CLI
 #   HEARTBEAT_INTERVAL    Heartbeat log interval in seconds (default: 30)
+#   STALE_KILL_THRESHOLD   Auto-kill session after N seconds of no progress (default: 900)
 #   HEARTBEAT_STALE_THRESHOLD  Heartbeat stale threshold in seconds (default: 600)
 #   LOG_CLEANUP_ENABLED   Run periodic log cleanup (default: 1)
 #   LOG_RETENTION_DAYS    Delete logs older than N days (default: 14)
@@ -40,6 +41,7 @@ MAX_RETRIES=${MAX_RETRIES:-3}
 SESSION_TIMEOUT=${SESSION_TIMEOUT:-0}
 HEARTBEAT_STALE_THRESHOLD=${HEARTBEAT_STALE_THRESHOLD:-600}
 HEARTBEAT_INTERVAL=${HEARTBEAT_INTERVAL:-30}
+STALE_KILL_THRESHOLD=${STALE_KILL_THRESHOLD:-900}
 LOG_CLEANUP_ENABLED=${LOG_CLEANUP_ENABLED:-1}
 LOG_RETENTION_DAYS=${LOG_RETENTION_DAYS:-14}
 LOG_MAX_TOTAL_MB=${LOG_MAX_TOTAL_MB:-1024}
@@ -148,8 +150,8 @@ spawn_and_wait_session() {
         watcher_pid=$!
     fi
 
-    # Heartbeat monitor
-    start_heartbeat "$cli_pid" "$session_log" "$progress_json" "$HEARTBEAT_INTERVAL"
+    # Heartbeat monitor (with stale-kill protection)
+    start_heartbeat "$cli_pid" "$session_log" "$progress_json" "$HEARTBEAT_INTERVAL" "$STALE_KILL_THRESHOLD"
     local heartbeat_pid="${_HEARTBEAT_PID:-}"
 
     # Wait for AI CLI to finish
@@ -168,6 +170,14 @@ spawn_and_wait_session() {
 
     [[ $exit_code -eq 143 ]] && exit_code=124
 
+    # Check for stale-kill marker (heartbeat killed the process due to no progress)
+    local stale_kill_marker="$session_dir/logs/stale-kill.json"
+    local was_stale_killed=false
+    if [[ -f "$stale_kill_marker" ]]; then
+        was_stale_killed=true
+        log_warn "Session was stale-killed by heartbeat monitor (no progress for too long)"
+    fi
+
     # Session summary
     if [[ -f "$session_log" ]]; then
         local final_size=$(wc -c < "$session_log" 2>/dev/null | tr -d ' ')
@@ -185,6 +195,33 @@ spawn_and_wait_session() {
     if [[ $exit_code -eq 124 ]]; then
         log_warn "Session timed out after ${SESSION_TIMEOUT}s"
         session_status="timed_out"
+    elif [[ "$was_stale_killed" == true ]]; then
+        log_warn "Session stale-killed (no progress for ${STALE_KILL_THRESHOLD}s)"
+        local has_commits=""
+        if git -C "$project_root" rev-parse --is-inside-work-tree >/dev/null 2>&1; then
+            has_commits=$(git -C "$project_root" log "${default_branch}..HEAD" --oneline 2>/dev/null | head -1)
+        fi
+        if [[ -n "$has_commits" ]]; then
+            log_info "Stale-killed session has commits — treating as success"
+            session_status="success"
+        else
+            local uncommitted=""
+            uncommitted=$(git -C "$project_root" status --porcelain 2>/dev/null | head -1 || true)
+            if [[ -n "$uncommitted" ]]; then
+                log_warn "Stale-killed session has uncommitted changes — auto-committing..."
+                git -C "$project_root" add -A 2>/dev/null || true
+                if git -C "$project_root" commit --no-verify -m "chore($refactor_id): auto-commit session work (stale-killed)" 2>/dev/null; then
+                    log_info "Auto-commit succeeded"
+                    session_status="success"
+                else
+                    log_warn "Auto-commit failed — no changes to commit"
+                    session_status="crashed"
+                fi
+            else
+                log_warn "Stale-killed session produced no commits and no changes"
+                session_status="crashed"
+            fi
+        fi
     elif [[ $exit_code -ne 0 ]]; then
         log_warn "Session exited with code $exit_code"
         session_status="crashed"
@@ -1129,6 +1166,7 @@ show_help() {
     echo "  VERBOSE                  Set to 1 for verbose AI CLI output"
     echo "  STRICT_BEHAVIOR_CHECK    Force full test suite after each refactor (default: 1)"
     echo "  HEARTBEAT_INTERVAL       Heartbeat log interval in seconds (default: 30)"
+    echo "  STALE_KILL_THRESHOLD     Auto-kill session after N seconds of no progress (default: 900)"
     echo "  LOG_CLEANUP_ENABLED      Run log cleanup before execution (default: 1)"
     echo "  LOG_RETENTION_DAYS       Delete logs older than N days (default: 14)"
     echo "  LOG_MAX_TOTAL_MB         Keep total logs under N MB (default: 1024)"

package/bundled/dev-pipeline/scripts/detect-stuck.py

@@ -226,15 +226,16 @@ def check_stuck_checkpoint(item_dir):
     return None
 
 
-def check_stale_heartbeat(item_id, item_status, state_dir, items_subdir, stale_threshold):
+def check_stale_heartbeat(item_id, item_status, state_dir, items_subdir, stale_threshold, task_list_status=None):
     """Check 3: Is the heartbeat stale or missing for an in_progress item?
 
     Only applies to items whose status indicates active work.
-    Uses last_session_id from the item's own status to find the active session.
+    Status is read from task_list_status (task list JSON, single source of truth).
+    Uses last_session_id from the item's own status.json to find the active session.
 
     Returns a stuck-report dict or None.
     """
-    status = item_status.get("status")
+    status = task_list_status
     # All pipelines now use "in_progress" as the active status
     in_progress_statuses = {"in_progress"}
     if status not in in_progress_statuses:
@@ -287,6 +288,8 @@ def check_stale_heartbeat(item_id, item_status, state_dir, items_subdir, stale_t
 def check_dependency_deadlock(item_id, task_list_data, state_dir, items_subdir, items_key):
     """Check 4: Does this item depend on a failed item?
 
+    Reads dependency status from task list JSON (single source of truth).
+
     Returns a stuck-report dict or None.
     """
     if task_list_data is None:
@@ -296,6 +299,12 @@ def check_dependency_deadlock(item_id, task_list_data, state_dir, items_subdir,
     if not isinstance(items, list):
         return None
 
+    # Build status map from task list
+    status_map = {}
+    for item in items:
+        if isinstance(item, dict) and item.get("id"):
+            status_map[item["id"]] = item.get("status", "pending")
+
     # Find this item in the list to get its dependencies
     deps = None
     for item in items:
@@ -308,15 +317,9 @@ def check_dependency_deadlock(item_id, task_list_data, state_dir, items_subdir,
     if not deps or not isinstance(deps, list):
         return None
 
-    # Check each dependency's status in state
+    # Check each dependency's status from the task list
     for dep_id in deps:
-        dep_status_path = os.path.join(
-            state_dir, items_subdir, dep_id, "status.json"
-        )
-        dep_status = load_json(dep_status_path)
-        if dep_status is None:
-            continue
-        dep_state = dep_status.get("status")
+        dep_state = status_map.get(dep_id)
         if dep_state == "failed":
             return {
                 "reason": "dependency_failed",
@@ -376,8 +379,16 @@ def check_item(item_id, state_dir, items_subdir, items_key, task_list_data, max_
     item_status = load_json(status_path)
 
     if item_status is None:
-        # Cannot read status — skip silently
-        return []
+        # Create a minimal runtime dict so checks can proceed
+        item_status = {}
+
+    # Look up item status from task list (single source of truth)
+    task_list_status = None
+    if task_list_data:
+        for item in task_list_data.get(items_key, []):
+            if isinstance(item, dict) and item.get("id") == item_id:
+                task_list_status = item.get("status", "pending")
+                break
 
     reports = []
 
@@ -392,7 +403,7 @@ def check_item(item_id, state_dir, items_subdir, items_key, task_list_data, max_
         reports.append(result)
 
     # Check 3: Stale heartbeat
-    result = check_stale_heartbeat(item_id, item_status, state_dir, items_subdir, stale_threshold)
+    result = check_stale_heartbeat(item_id, item_status, state_dir, items_subdir, stale_threshold, task_list_status)
     if result is not None:
         reports.append(result)
 

package/bundled/dev-pipeline/scripts/init-bugfix-pipeline.py

@@ -249,13 +249,8 @@ def create_state_directory(state_dir, bug_list_path, bugs):
         sessions_dir = os.path.join(bug_dir, "sessions")
         os.makedirs(sessions_dir, exist_ok=True)
 
-        # Respect existing terminal status from bug-fix-list.json
-        bl_status = bug.get("status", "pending")
-        init_status = bl_status if bl_status in TERMINAL_STATUSES else "pending"
-
         bug_status = {
             "bug_id": bid,
-            "status": init_status,
             "retry_count": 0,
             "max_retries": 3,
             "sessions": [],