fix(loop): guard bare function calls against set -e script termination

CLAUDE.md

@@ -45,13 +45,14 @@ The system uses a modular architecture with reusable components in the `lib/` di
    - Analyzes Claude Code output for completion signals
    - **JSON output format detection and parsing** (with text fallback)
    - Supports both flat JSON format and Claude CLI format (`result`, `sessionId`, `metadata`)
-   - Extracts structured fields: status, exit_signal, work_type, files_modified
+   - Extracts structured fields: status, exit_signal, work_type, files_modified, asking_questions, question_count
+   - **Question detection**: `detect_questions()` with `QUESTION_PATTERNS` array — detects when Claude asks questions instead of acting autonomously (Issue #190)
    - **Session management**: `store_session_id()`, `get_last_session_id()`, `should_resume_session()`
    - Automatic session persistence to `.ralph/.claude_session_id` file with 24-hour expiration
    - Session lifecycle: `get_session_id()`, `reset_session()`, `log_session_transition()`, `init_session_tracking()`
    - Session history tracked in `.ralph/.ralph_session_history` (last 50 transitions)
    - Session auto-reset on: circuit breaker open, manual interrupt, project completion
-   - Detects test-only loops and stuck error patterns
+   - Detects test-only loops, stuck error patterns, and question-only loops
    - Two-stage error filtering to eliminate false positives
    - Multi-line error matching for accurate stuck loop detection
    - Confidence scoring for exit decisions
@@ -176,7 +177,7 @@ tmux attach -t <session-name>
 
 ### Running Tests
 ```bash
-# Run all tests (568 tests)
+# Run all tests (554 tests)
 npm test
 
 # Run specific test suites
@@ -221,13 +222,23 @@ CLAUDE_OUTPUT_FORMAT="json"           # Output format: json (default) or text
 CLAUDE_ALLOWED_TOOLS="Write,Read,Edit,Bash(git add *),Bash(git commit *),...,Bash(npm *),Bash(pytest)"  # Allowed tool permissions (see File Protection)
 CLAUDE_USE_CONTINUE=true              # Enable session continuity
 CLAUDE_MIN_VERSION="2.0.76"           # Minimum Claude CLI version
+CLAUDE_AUTO_UPDATE=true               # Auto-update Claude CLI at startup (set false for air-gapped environments)
 ```
 
+**Auto-Update Configuration:**
+- `CLAUDE_AUTO_UPDATE` controls whether Ralph checks npm registry and attempts `npm update -g` at startup
+- **Local workstation / home server**: Keep `true` (default) — CLI updates include bug fixes and new features that improve Ralph's effectiveness. The 200-500ms startup overhead is negligible for loops that run hours
+- **Docker container**: Set `false` in `.ralphrc` — container is ephemeral and version is pinned at image build time. The npm registry query and potential update are pure overhead
+- **Air-gapped environment**: Set `false` — npm registry is unreachable, the check will timeout and log a warning
+- Update failure is non-blocking: Ralph logs a warning and continues the loop normally
+
 **Claude Code CLI Command (Issue #97):**
 - `CLAUDE_CODE_CMD` defaults to `"claude"` (global install)
 - Configurable via `.ralphrc` for alternative installations (e.g., `"npx @anthropic-ai/claude-code"`)
 - Auto-detected during `ralph-enable` and `ralph-setup` (prefers `claude` if available, falls back to npx)
 - Validated at startup with `validate_claude_command()` — displays clear error with installation instructions if not found
+- After validation, `check_claude_version()` verifies minimum version compatibility and `check_claude_updates()` queries npm registry for latest version with auto-update attempt (Issue #190)
+- Both functions use `compare_semver()` for proper major→minor→patch sequential comparison (safe for any patch number, unlike integer arithmetic)
 - Environment variable `CLAUDE_CODE_CMD` takes precedence over `.ralphrc`
 
 **CLI Options:**
@@ -241,6 +252,7 @@ Each loop iteration injects context via `build_loop_context()`:
 - Remaining tasks from fix_plan.md
 - Circuit breaker state (if not CLOSED)
 - Previous loop work summary
+- Corrective guidance if previous loop detected questions (Issue #190)
 
 **Session Continuity:**
 - Sessions are preserved in `.ralph/.claude_session_id`
@@ -408,6 +420,8 @@ When Claude Code exceeds `CLAUDE_TIMEOUT_MINUTES`, `portable_timeout` terminates
 
 In both modes, a timeout results in `exit_code=124`, which flows into the standard error handling path (logged, circuit breaker updated, loop continues to next iteration).
 
+**Stderr separation (Issue #190):** In live mode, Claude CLI stderr (e.g., Node.js UNDICI warnings) is redirected to a separate file (`claude_stderr_<timestamp>.log`) instead of being merged into the stdout JSON stream via `2>&1`. This prevents non-JSON stderr content from corrupting the `jq` pipeline, which previously caused jq exit code 4, empty `live.log`, and no terminal output. Background mode is unaffected (stderr goes directly to the log file). If stderr content is detected, a WARN is logged referencing the stderr file.
+
 ### API Limit Detection (Issue #183)
 
 The API 5-hour limit detection uses a three-layer approach to avoid false positives. In stream-json mode, output files contain echoed file content from tool results (`"type":"user"` lines). If project files mention "5-hour limit", naive grep patterns match those echoed strings, incorrectly triggering the API limit recovery flow.
@@ -428,6 +442,7 @@ Only searches `tail -30` of the output file, filtering out `"type":"user"`, `"to
 - `CB_SAME_ERROR_THRESHOLD=5` - Open circuit after 5 loops with repeated errors
 - `CB_OUTPUT_DECLINE_THRESHOLD=70%` - Open circuit if output declines by >70%
 - `CB_PERMISSION_DENIAL_THRESHOLD=2` - Open circuit after 2 loops with permission denials (Issue #101)
+- **Question loop suppression** (Issue #190): When `asking_questions=true`, the `consecutive_no_progress` counter is held steady (not incremented). This prevents the circuit breaker from opening prematurely when Claude asks questions in headless mode. A corrective message is injected via `build_loop_context()` in the next loop iteration.
 
 ### Circuit Breaker Auto-Recovery (Issue #160)
 
@@ -519,17 +534,17 @@ Ralph uses a multi-layered strategy to prevent Claude from accidentally deleting
 
 ## Test Suite
 
-### Test Files (568 tests total)
+### Test Files (559 tests total)
 
 | File | Tests | Description |
 |------|-------|-------------|
-| `test_circuit_breaker_recovery.bats` | 19 | Cooldown timer, auto-reset, parse_iso_to_epoch, CLI flag (Issue #160) |
+| `test_circuit_breaker_recovery.bats` | 20 | Cooldown timer, auto-reset, parse_iso_to_epoch, CLI flag (Issue #160) + log_circuit_transition jq resilience |
 | `test_cli_parsing.bats` | 35 | CLI argument parsing for all flags + monitor parameter forwarding |
-| `test_cli_modern.bats` | 68 | Modern CLI commands (Phase 1.1) + build_claude_command fix + live mode text format fix (#164) + errexit pipeline guard (#175) + ALLOWED_TOOLS tightening (#149) + API limit false positive detection (#183) + Claude CLI command validation (#97) + stale call counter fix (#196) |
-| `test_json_parsing.bats` | 52 | JSON output format parsing + Claude CLI format + session management + array format |
-| `test_session_continuity.bats` | 44 | Session lifecycle management + expiration + circuit breaker integration + issue #91 fix |
-| `test_exit_detection.bats` | 53 | Exit signal detection + EXIT_SIGNAL-based completion indicators + progress detection |
-| `test_rate_limiting.bats` | 15 | Rate limiting behavior |
+| `test_cli_modern.bats` | 92 | Modern CLI commands (Phase 1.1) + build_claude_command fix + live mode text format fix (#164) + errexit pipeline guard (#175) + ALLOWED_TOOLS tightening (#149) + API limit false positive detection (#183) + Claude CLI command validation (#97) + stale call counter fix (#196) + question detection corrective message (#190) + stderr separation (#190) + version check and auto-update at startup (#190) + semver comparison (#190) + set -e bare call safety (#190) |
+| `test_json_parsing.bats` | 50 | JSON output format parsing + Claude CLI format + array format + question detection (#190) |
+| `test_session_continuity.bats` | 26 | Session lifecycle management + expiration + circuit breaker integration + issue #91 fix |
+| `test_exit_detection.bats` | 50 | Exit signal detection + EXIT_SIGNAL-based completion indicators + progress detection + question detection integration (#190) |
+| `test_rate_limiting.bats` | 11 | Rate limiting behavior |
 | `test_loop_execution.bats` | 20 | Integration tests |
 | `test_edge_cases.bats` | 25 | Edge case handling |
 | `test_installation.bats` | 15 | Global installation/uninstall workflows + dotfile template copying (#174) |
@@ -539,8 +554,8 @@ Ralph uses a multi-layered strategy to prevent Claude from accidentally deleting
 | `test_task_sources.bats` | 23 | Task sources (beads, GitHub, PRD extraction, normalization) |
 | `test_ralph_enable.bats` | 24 | Ralph enable integration tests (wizard, CI version, JSON output, .ralphrc validation #149) |
 | `test_wizard_utils.bats` | 20 | Wizard utility functions (stdout/stderr separation, prompt functions) |
-| `test_file_protection.bats` | 22 | File integrity validation (RALPH_REQUIRED_PATHS, validate_ralph_integrity, get_integrity_report) (Issue #149) |
-| `test_integrity_check.bats` | 12 | Pre-loop integrity check in ralph_loop.sh (startup + in-loop validation) (Issue #149) |
+| `test_file_protection.bats` | 15 | File integrity validation (RALPH_REQUIRED_PATHS, validate_ralph_integrity, get_integrity_report) (Issue #149) |
+| `test_integrity_check.bats` | 10 | Pre-loop integrity check in ralph_loop.sh (startup + in-loop validation) (Issue #149) |
 
 ### Running Tests
 ```bash

lib/circuit_breaker.sh

@@ -206,6 +206,12 @@ record_loop_result() {
         consecutive_permission_denials=0
     fi
 
+    # Check if Claude is asking questions (Issue #190 Bug 2)
+    local asking_questions="false"
+    if [[ -f "$response_analysis_file" ]]; then
+        asking_questions=$(jq -r '.analysis.asking_questions // false' "$response_analysis_file" 2>/dev/null || echo "false")
+    fi
+
     # Determine if progress was made
     if [[ $files_changed -gt 0 ]]; then
         # Git shows uncommitted changes - clear progress
@@ -223,6 +229,10 @@ record_loop_result() {
         has_progress=true
         consecutive_no_progress=0
         last_progress_loop=$loop_number
+    elif [[ "$asking_questions" == "true" ]]; then
+        # Claude is asking questions — not progress, but not stagnation either.
+        # Suppress no-progress counter; corrective context will redirect next loop.
+        has_progress=false
     else
         consecutive_no_progress=$((consecutive_no_progress + 1))
     fi
@@ -340,8 +350,16 @@ log_circuit_transition() {
         \"reason\": \"$reason\"
     }"
 
-    history=$(echo "$history" | jq ". += [$transition]")
-    echo "$history" > "$CB_HISTORY_FILE"
+    local updated_history
+    updated_history=$(echo "$history" | jq ". += [$transition]" 2>/dev/null)
+    local jq_status=$?
+
+    if [[ $jq_status -eq 0 && -n "$updated_history" ]]; then
+        echo "$updated_history" > "$CB_HISTORY_FILE"
+    else
+        # Fallback: preserve current transition only (history file was corrupted)
+        echo "[$transition]" > "$CB_HISTORY_FILE"
+    fi
 
     # Console log with colors
     case $to_state in

lib/enable_core.sh

@@ -728,6 +728,9 @@ BEADS_FILTER="status:open"
 CB_NO_PROGRESS_THRESHOLD=3
 CB_SAME_ERROR_THRESHOLD=5
 CB_OUTPUT_DECLINE_THRESHOLD=70
+
+# Auto-update Claude CLI at startup
+CLAUDE_AUTO_UPDATE=true
 RALPHRCEOF
 }
 

lib/response_analyzer.sh

@@ -22,6 +22,33 @@ RALPH_DIR="${RALPH_DIR:-.ralph}"
 COMPLETION_KEYWORDS=("done" "complete" "finished" "all tasks complete" "project complete" "ready for review")
 TEST_ONLY_PATTERNS=("npm test" "bats" "pytest" "jest" "cargo test" "go test" "running tests")
 NO_WORK_PATTERNS=("nothing to do" "no changes" "already implemented" "up to date")
+QUESTION_PATTERNS=("should I" "would you" "do you want" "which approach" "which option" "how should" "what should" "shall I" "do you prefer" "can you clarify" "could you" "what do you think" "please confirm" "need clarification" "awaiting.*input" "waiting.*response" "your preference")
+
+# Detect if Claude is asking questions instead of acting autonomously
+# Args: $1 = text content to analyze
+# Returns: 0 if questions detected, 1 otherwise
+# Outputs: question count on stdout
+detect_questions() {
+    local content="$1"
+    local question_count=0
+
+    if [[ -z "$content" ]]; then
+        echo "0"
+        return 1
+    fi
+
+    # Count lines matching question patterns (case-insensitive)
+    for pattern in "${QUESTION_PATTERNS[@]}"; do
+        local matches
+        matches=$(echo "$content" | grep -ciw "$pattern" 2>/dev/null || echo "0")
+        matches=$(echo "$matches" | tr -d '[:space:]')
+        matches=${matches:-0}
+        question_count=$((question_count + matches))
+    done
+
+    echo "$question_count"
+    [[ $question_count -gt 0 ]] && return 0 || return 1
+}
 
 # =============================================================================
 # JSON OUTPUT FORMAT DETECTION AND PARSING
@@ -359,6 +386,13 @@ analyze_response() {
                 confidence_score=$((json_confidence + 50))
             fi
 
+            # Detect questions in JSON response text (Issue #190 Bug 2)
+            local asking_questions=false
+            local question_count=0
+            if question_count=$(detect_questions "$work_summary"); then
+                asking_questions=true
+            fi
+
             # Check for file changes via git (supplements JSON data)
             # Fix #141: Detect both uncommitted changes AND committed changes
             if command -v git &>/dev/null && git rev-parse --git-dir >/dev/null 2>&1; then
@@ -415,6 +449,8 @@ analyze_response() {
                 --argjson has_permission_denials "$has_permission_denials" \
                 --argjson permission_denial_count "$permission_denial_count" \
                 --argjson denied_commands "$denied_commands_json" \
+                --argjson asking_questions "$asking_questions" \
+                --argjson question_count "$question_count" \
                 '{
                     loop_number: $loop_number,
                     timestamp: $timestamp,
@@ -432,7 +468,9 @@ analyze_response() {
                         output_length: $output_length,
                         has_permission_denials: $has_permission_denials,
                         permission_denial_count: $permission_denial_count,
-                        denied_commands: $denied_commands
+                        denied_commands: $denied_commands,
+                        asking_questions: $asking_questions,
+                        question_count: $question_count
                     }
                 }' > "$analysis_result_file"
             rm -f "$RALPH_DIR/.json_parse_result"
@@ -530,6 +568,14 @@ analyze_response() {
         fi
     done
 
+    # 5.5. Detect question patterns (Claude asking instead of acting) (Issue #190 Bug 2)
+    local asking_questions=false
+    local question_count=0
+    if question_count=$(detect_questions "$output_content"); then
+        asking_questions=true
+        work_summary="Claude is asking questions instead of acting autonomously"
+    fi
+
     # 6. Check for file changes (git integration)
     # Fix #141: Detect both uncommitted changes AND committed changes
     if command -v git &>/dev/null && git rev-parse --git-dir >/dev/null 2>&1; then
@@ -613,6 +659,8 @@ analyze_response() {
         --argjson exit_signal "$exit_signal" \
         --arg work_summary "$work_summary" \
         --argjson output_length "$output_length" \
+        --argjson asking_questions "$asking_questions" \
+        --argjson question_count "$question_count" \
         '{
             loop_number: $loop_number,
             timestamp: $timestamp,
@@ -630,7 +678,9 @@ analyze_response() {
                 output_length: $output_length,
                 has_permission_denials: false,
                 permission_denial_count: 0,
-                denied_commands: []
+                denied_commands: [],
+                asking_questions: $asking_questions,
+                question_count: $question_count
             }
         }' > "$analysis_result_file"
 
@@ -879,6 +929,7 @@ export -f analyze_response
 export -f update_exit_signals
 export -f log_analysis_summary
 export -f detect_stuck_loop
+export -f detect_questions
 export -f store_session_id
 export -f get_last_session_id
 export -f should_resume_session