loki-mode 5.49.0 → 5.49.1

package/README.md

@@ -1,6 +1,6 @@
 # Loki Mode
 
-**The Flagship Product of [Autonomi](https://www.autonomi.dev/) -- The First Truly Autonomous Multi-Agent Startup System**
+**The Flagship Product of [Autonomi](https://www.autonomi.dev/) -- An Autonomous Multi-Agent Development System**
 
 [![npm version](https://img.shields.io/npm/v/loki-mode)](https://www.npmjs.com/package/loki-mode)
 [![npm downloads](https://img.shields.io/npm/dw/loki-mode)](https://www.npmjs.com/package/loki-mode)
@@ -9,17 +9,15 @@
 [![GitHub Marketplace](https://img.shields.io/badge/Marketplace-Loki%20Mode-purple?logo=github)](https://github.com/marketplace/actions/loki-mode-code-review)
 [![Autonomi](https://img.shields.io/badge/Autonomi-autonomi.dev-5B4EEA)](https://www.autonomi.dev/)
 [![Agent Types](https://img.shields.io/badge/Agent%20Types-41-blue)]()
-[![Loki Mode](https://img.shields.io/badge/Loki%20Mode-98.78%25%20Pass%401-blueviolet)](benchmarks/results/)
-[![HumanEval](https://img.shields.io/badge/HumanEval-98.17%25%20Pass%401-brightgreen)](benchmarks/results/)
-[![SWE-bench](https://img.shields.io/badge/SWE--bench-99.67%25%20Patch%20Gen-brightgreen)](benchmarks/results/)
+[![Benchmarks](https://img.shields.io/badge/Benchmarks-Infrastructure%20Ready-blue)](benchmarks/)
 
-**Current Version: v5.47.0**
+**Current Version: v5.49.0**
 
 **[Autonomi](https://www.autonomi.dev/)** | **[Documentation](https://www.autonomi.dev/docs)** | **[GitHub](https://github.com/asklokesh/loki-mode)**
 
-> **PRD → Deployed Product in Zero Human Intervention**
+> **PRD to Deployed Product with Minimal Human Intervention**
 >
-> Loki Mode transforms a Product Requirements Document into a fully built, tested, deployed, and revenue-generating product while you sleep. No manual steps. No intervention. Just results.
+> Loki Mode transforms a Product Requirements Document into a fully built, tested, and deployed product with autonomous multi-agent execution. Human oversight for deployment credentials, domain setup, and critical decisions.
 
 ---
 
@@ -79,7 +77,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
-      - uses: asklokesh/loki-mode@v5.38
+      - uses: asklokesh/loki-mode@v5
         with:
           github_token: ${{ secrets.GITHUB_TOKEN }}
           mode: review          # review, fix, or test
@@ -163,40 +161,27 @@ See [skills/providers.md](skills/providers.md) for full provider documentation.
 
 ---
 
-## Benchmark Results
+## Benchmarks
 
-### Three-Way Comparison (HumanEval)
+Benchmark infrastructure is included for HumanEval and SWE-bench evaluation. Results are self-reported from the included test harness and have not been independently verified.
 
-| System | Pass@1 | Details |
-|--------|--------|---------|
-| **Loki Mode (Multi-Agent)** | **98.78%** | 162/164 problems, RARV cycle recovered 2 |
-| Direct Claude | 98.17% | 161/164 problems (baseline) |
-| MetaGPT | 85.9-87.7% | Published benchmark |
+| Benchmark | Result | Notes |
+|-----------|--------|-------|
+| HumanEval | 162/164 (98.78%) | Self-reported, max 3 retries per problem |
+| SWE-bench | 299/300 patches generated | Patch generation only -- SWE-bench evaluator not yet run to verify correctness |
 
-**Loki Mode beats MetaGPT by +11-13%** thanks to the RARV (Reason-Act-Reflect-Verify) cycle.
+**Note:** SWE-bench "patch generation" means the system produced a patch file, not that the patch correctly resolves the issue. The SWE-bench evaluator should be run to determine actual resolution rates.
 
-### Full Results
-
-| Benchmark | Score | Details |
-|-----------|-------|---------|
-| **Loki Mode HumanEval** | **98.78% Pass@1** | 162/164 (multi-agent with RARV) |
-| **Direct Claude HumanEval** | **98.17% Pass@1** | 161/164 (single agent baseline) |
-| **Direct Claude SWE-bench** | **99.67% patch gen** | 299/300 problems |
-| **Loki Mode SWE-bench** | **99.67% patch gen** | 299/300 problems |
-| Model | Claude Opus 4.5 | |
-
-**Key Finding:** Multi-agent RARV matches single-agent performance on both benchmarks after timeout optimization. The 4-agent pipeline (Architect->Engineer->QA->Reviewer) achieves the same 99.67% patch generation as direct Claude.
-
-See [benchmarks/results/](benchmarks/results/) for full methodology and solutions.
+See [benchmarks/](benchmarks/) for the test harness and raw results.
 
 ---
 
 ## What is Loki Mode?
 
-Loki Mode is a multi-provider AI skill that orchestrates **41 specialized AI agent types** across **7 swarms** to autonomously build, test, deploy, and scale complete startups. Works with **Claude Code**, **OpenAI Codex CLI**, and **Google Gemini CLI**. It dynamically spawns only the agents you need—**5-10 for simple projects, 100+ for complex startups**—working in parallel with continuous self-verification.
+Loki Mode is a multi-provider AI skill that orchestrates **41 specialized AI agent types** across **8 swarms** to autonomously build, test, and deploy software projects. Works with **Claude Code**, **OpenAI Codex CLI**, and **Google Gemini CLI**. It dynamically spawns agents as needed -- typically **5-10 for simple projects, more for complex ones** -- working in parallel with continuous self-verification.
 
 ```
-PRD → Research → Architecture → Development → Testing → Deployment → Marketing → Revenue
+PRD → Research → Architecture → Development → Testing → Deployment → Marketing
 ```
 
 **Just say "Loki Mode" and point to a PRD. Walk away. Come back to a deployed product.**
@@ -205,11 +190,11 @@ PRD → Research → Architecture → Development → Testing → Deployment →
 
 ## Why Loki Mode?
 
-### **Better Than Anything Out There**
+### **How It Works**
 
 | What Others Do | What Loki Mode Does |
 |----------------|---------------------|
-| **Single agent** writes code linearly | **100+ agents** work in parallel across engineering, ops, business, data, product, and growth |
+| **Single agent** writes code linearly | **Multiple agents** work in parallel across engineering, ops, business, data, product, and growth |
 | **Manual deployment** required | **Autonomous deployment** to AWS, GCP, Azure, Vercel, Railway with blue-green and canary strategies |
 | **No testing** or basic unit tests | **7 automated quality gates**: input/output guardrails, static analysis, blind review, anti-sycophancy, severity blocking, test coverage |
 | **Code only** - you handle the rest | **Full business operations**: marketing, sales, legal, HR, finance, investor relations |
@@ -221,8 +206,8 @@ PRD → Research → Architecture → Development → Testing → Deployment →
 
 ### **Core Advantages**
 
-1. **Truly Autonomous**: RARV (Reason-Act-Reflect-Verify) cycle with self-verification achieves 2-3x quality improvement
-2. **Massively Parallel**: 100+ agents working simultaneously, not sequential single-agent bottlenecks
+1. **Self-Verifying**: RARV (Reason-Act-Reflect-Verify) cycle with continuous self-verification catches errors early
+2. **Parallel Execution**: Multiple agents working simultaneously, not sequential single-agent bottlenecks
 3. **Production-Ready**: Not just code—handles deployment, monitoring, incident response, and business operations
 4. **Self-Improving**: Learns from mistakes, updates continuity logs, prevents repeated errors
 5. **Zero Babysitting**: Auto-resumes on rate limits, recovers from failures, runs until completion
@@ -255,7 +240,7 @@ PRD → Research → Architecture → Development → Testing → Deployment →
 | **GitHub Integration** | Issue import, PR creation, status sync | [GitHub Integration](skills/github-integration.md) |
 | **Distribution** | npm, Homebrew, Docker installation | [Installation Guide](docs/INSTALLATION.md) |
 | **Research Foundation** | OpenAI, DeepMind, Anthropic patterns | [Acknowledgements](docs/ACKNOWLEDGEMENTS.md) |
-| **Benchmarks** | HumanEval 98.78%, SWE-bench 99.67% | [Benchmark Results](benchmarks/results/) |
+| **Benchmarks** | HumanEval and SWE-bench infrastructure included | [Benchmark Harness](benchmarks/) |
 | **Comparisons** | vs Auto-Claude, Cursor | [Auto-Claude](docs/auto-claude-comparison.md), [Cursor](docs/cursor-comparison.md) |
 
 ---
@@ -424,7 +409,7 @@ Loki Mode doesn't just write code—it **thinks, acts, learns, and verifies**:
    └─ Apply learning and RETRY from REASON
 ```
 
-**Result:** 2-3x quality improvement through continuous self-verification.
+**Result:** Improved quality through continuous self-verification and multi-reviewer code review.
 
 ### **Perpetual Improvement Mode**
 
@@ -561,7 +546,7 @@ graph TB
 **Key components:**
 - **RARV+C Cycle** -- Reason, Act, Reflect, Verify, Compound. Every iteration follows this loop. Failed verification triggers retry from Reason.
 - **Provider Layer** -- Claude Code (full parallel agents, Task tool, MCP), Codex CLI and Gemini CLI (sequential, degraded mode).
-- **Agent Swarms** -- 41 specialized agent types across 7 swarms, spawned on demand based on project complexity.
+- **Agent Swarms** -- 41 specialized agent types across 8 swarms, spawned on demand based on project complexity.
 - **Completion Council** -- 3 members vote on whether the project is done. Anti-sycophancy devil's advocate on unanimous votes.
 - **Memory System** -- Episodic traces, semantic patterns, procedural skills. Progressive disclosure reduces context usage by 60-80%.
 - **Dashboard** -- FastAPI server reading `.loki/` flat files, with real-time web UI for task queue, agents, logs, and council state. Now with TLS/HTTPS, OIDC/SSO, and RBAC (v5.36.0-v5.37.0).
@@ -609,7 +594,7 @@ Config search order: `.loki/config.yaml` (project) -> `~/.config/loki-mode/confi
 
 ## Agent Swarms (41 Types)
 
-Loki Mode has **41 predefined agent types** organized into **7 specialized swarms**. The orchestrator spawns only what you need—simple projects use 5-10 agents, complex startups spawn 100+.
+Loki Mode has **41 predefined agent types** organized into **8 specialized swarms**. The orchestrator spawns only what you need -- simple projects typically use 5-10 agents, complex ones may use more.
 
 <img width="5309" height="979" alt="Agent Swarms Visualization" src="https://github.com/user-attachments/assets/7d18635d-a606-401f-8d9f-430e6e4ee689" />
 
@@ -981,7 +966,7 @@ Built for the [Claude Code](https://claude.ai) ecosystem, powered by Anthropic's
 
 Loki Mode is the flagship product of **[Autonomi](https://www.autonomi.dev/)** -- a platform for autonomous AI systems. Like Alphabet is to Google, Autonomi is the parent brand under which Loki Mode and future products operate.
 
-**Why Autonomi?** Loki Mode proved that multi-agent autonomous systems can build real software from a PRD with zero human intervention. Autonomi is the expansion of that vision into a broader platform of autonomous services and products.
+**Why Autonomi?** Loki Mode proved that multi-agent autonomous systems can build real software from a PRD with minimal human intervention. Autonomi is the expansion of that vision into a broader platform of autonomous services and products.
 
 - **[autonomi.dev](https://www.autonomi.dev/)** -- Main website
 - **[Documentation](https://www.autonomi.dev/docs)** -- Full documentation

package/SKILL.md

@@ -1,9 +1,9 @@
 ---
 name: loki-mode
-description: Multi-agent autonomous startup system. Triggers on "Loki Mode". Takes PRD to deployed product with zero human intervention. Requires --dangerously-skip-permissions flag.
+description: Multi-agent autonomous startup system. Triggers on "Loki Mode". Takes PRD to deployed product with minimal human intervention. Requires --dangerously-skip-permissions flag.
 ---
 
-# Loki Mode v5.49.0
+# Loki Mode v5.49.1
 
 **You are an autonomous agent. You make decisions. You do not ask questions. You do not stop.**
 
@@ -263,4 +263,4 @@ The following features are documented in skill modules but not yet fully automat
 | Quality gates 3-reviewer system | Implemented (v5.35.0) | 5 specialist reviewers in `skills/quality-gates.md`; execution in run.sh |
 | Benchmarks (HumanEval, SWE-bench) | Infrastructure only | Runner scripts and datasets exist in `benchmarks/`; no published results |
 
-**v5.49.0 | [Autonomi](https://www.autonomi.dev/) flagship product | ~260 lines core**
+**v5.49.1 | [Autonomi](https://www.autonomi.dev/) flagship product | ~260 lines core**

package/VERSION

@@ -1 +1 @@
-5.49.0
+5.49.1

package/autonomy/CONSTITUTION.md

@@ -142,7 +142,7 @@ GROWTH ──[continuous improvement loop]──> GROWTH
 - `Bash` - Command execution
 - `platform-orchestrator` - Deployment and service management
 
-**The 37 agent types are ROLES defined through prompts, not subagent_types.**
+**The 41 agent types are ROLES defined through prompts, not subagent_types.**
 
 ---
 
@@ -158,7 +158,7 @@ skills/
   quality-gates.md             # 7-gate system, anti-sycophancy
   testing.md                   # Playwright, E2E, property-based
   production.md                # CI/CD, batch processing
-  agents.md                    # 37 agent types, A2A patterns
+  agents.md                    # 41 agent types, A2A patterns
   parallel-workflows.md        # Git worktrees, parallel streams
   troubleshooting.md           # Error recovery, fallbacks
   artifacts.md                 # Code generation patterns

package/autonomy/app-runner.sh

@@ -432,6 +432,10 @@ app_runner_start() {
         (cd "$dir" && bash -c "$_APP_RUNNER_METHOD" >> "$_APP_RUNNER_DIR/app.log" 2>&1) &
     fi
     _APP_RUNNER_PID=$!
+    # Register with central PID registry if available
+    if type register_pid &>/dev/null; then
+        register_pid "$_APP_RUNNER_PID" "app-runner" "method=$_APP_RUNNER_METHOD"
+    fi
 
     # Write PID file
     echo "$_APP_RUNNER_PID" > "$_APP_RUNNER_DIR/app.pid"
@@ -497,6 +501,11 @@ app_runner_stop() {
         kill -KILL "-$_APP_RUNNER_PID" 2>/dev/null || kill -KILL "$_APP_RUNNER_PID" 2>/dev/null || true
     fi
 
+    # Unregister from central PID registry
+    if type unregister_pid &>/dev/null && [ -n "$_APP_RUNNER_PID" ]; then
+        unregister_pid "$_APP_RUNNER_PID"
+    fi
+
     rm -f "$_APP_RUNNER_DIR/app.pid"
     _write_app_state "stopped"
     log_info "App Runner: application stopped"

package/autonomy/loki

@@ -9,6 +9,7 @@
 # Usage:
 #   loki start [PRD]      - Start Loki Mode (optionally with PRD)
 #   loki stop             - Stop execution immediately
+#   loki cleanup          - Kill orphaned processes from crashed sessions
 #   loki pause            - Pause after current session
 #   loki resume           - Resume paused execution
 #   loki status           - Show current status
@@ -312,6 +313,7 @@ show_help() {
     echo "  init             Build a PRD interactively or from templates"
     echo "  issue <url|num>  Generate PRD from GitHub issue and optionally start"
     echo "  stop             Stop execution immediately"
+    echo "  cleanup          Kill orphaned processes from crashed sessions"
     echo "  pause            Pause after current session"
     echo "  resume           Resume paused execution"
     echo "  status [--json]  Show current status (--json for machine-readable)"
@@ -704,6 +706,28 @@ except: pass
             rm -f "$LOKI_DIR/dashboard/dashboard.pid"
         fi
 
+        # Kill any remaining registered processes (2s graceful window matches run.sh)
+        if [ -d "$LOKI_DIR/pids" ]; then
+            for entry_file in "$LOKI_DIR/pids"/*.json; do
+                [ -f "$entry_file" ] || continue
+                local reg_pid
+                reg_pid=$(basename "$entry_file" .json)
+                case "$reg_pid" in ''|*[!0-9]*) continue ;; esac
+                if kill -0 "$reg_pid" 2>/dev/null; then
+                    kill "$reg_pid" 2>/dev/null || true
+                    local w=0
+                    while [ $w -lt 4 ] && kill -0 "$reg_pid" 2>/dev/null; do
+                        sleep 0.5
+                        w=$((w + 1))
+                    done
+                    if kill -0 "$reg_pid" 2>/dev/null; then
+                        kill -9 "$reg_pid" 2>/dev/null || true
+                    fi
+                fi
+                rm -f "$entry_file"
+            done
+        fi
+
         # Emit session stop event
         emit_event session cli stop "reason=user_requested"
         # Emit success pattern for clean stop (SYN-018)
@@ -730,6 +754,86 @@ except: pass
     fi
 }
 
+# Kill orphaned processes from crashed sessions
+cmd_cleanup() {
+    local pids_dir="$LOKI_DIR/pids"
+    local killed=0
+    local stale=0
+
+    if [ ! -d "$pids_dir" ]; then
+        echo "No PID registry found. Nothing to clean up."
+        exit 0
+    fi
+
+    echo -e "${BOLD}Scanning for orphaned processes...${NC}"
+
+    for entry_file in "$pids_dir"/*.json; do
+        [ -f "$entry_file" ] || continue
+        local pid
+        pid=$(basename "$entry_file" .json)
+        case "$pid" in
+            ''|*[!0-9]*) continue ;;
+        esac
+
+        local label=""
+        local ppid_val=""
+        # Parse JSON fields (python3 with shell fallback)
+        if command -v python3 >/dev/null 2>&1; then
+            label=$(python3 -c "import json,sys; print(json.load(open(sys.argv[1])).get('label','unknown'))" "$entry_file" 2>/dev/null) || label="unknown"
+            ppid_val=$(python3 -c "import json,sys; print(json.load(open(sys.argv[1])).get('ppid',''))" "$entry_file" 2>/dev/null) || true
+        else
+            label=$(sed 's/.*"label":"//' "$entry_file" 2>/dev/null | sed 's/".*//' | head -1) || label="unknown"
+            ppid_val=$(sed 's/.*"ppid"://' "$entry_file" 2>/dev/null | sed 's/[,}].*//' | head -1) || true
+        fi
+
+        if kill -0 "$pid" 2>/dev/null; then
+            # Process is alive - check if parent is dead (orphan)
+            local is_orphan=false
+            # Validate ppid_val is numeric before using with kill
+            case "$ppid_val" in ''|*[!0-9]*) ppid_val="" ;; esac
+            if [ -n "$ppid_val" ] && ! kill -0 "$ppid_val" 2>/dev/null; then
+                is_orphan=true
+            fi
+
+            if [ "$is_orphan" = true ] || [ "${1:-}" = "--force" ]; then
+                echo -e "  ${RED}Killing${NC} PID=$pid label=$label (parent $ppid_val dead)"
+                kill "$pid" 2>/dev/null || true
+                sleep 0.5
+                if kill -0 "$pid" 2>/dev/null; then
+                    kill -9 "$pid" 2>/dev/null || true
+                fi
+                rm -f "$entry_file"
+                killed=$((killed + 1))
+            else
+                echo -e "  ${GREEN}Alive${NC}  PID=$pid label=$label (parent $ppid_val alive)"
+            fi
+        else
+            # Process is dead - clean up stale entry
+            rm -f "$entry_file"
+            stale=$((stale + 1))
+        fi
+    done
+
+    echo ""
+    echo "Results: $killed orphan(s) killed, $stale stale entries cleaned"
+
+    # Also kill orphaned loki-run temp scripts
+    local temp_killed=0
+    if pgrep -f "loki-run-" >/dev/null 2>&1; then
+        if ! is_session_running; then
+            echo "Killing orphaned loki-run temp scripts..."
+            pkill -f "loki-run-" 2>/dev/null || true
+            sleep 0.5
+            pkill -9 -f "loki-run-" 2>/dev/null || true
+            temp_killed=1
+        fi
+    fi
+
+    if [ $killed -eq 0 ] && [ $stale -eq 0 ] && [ $temp_killed -eq 0 ]; then
+        echo -e "${GREEN}System is clean. No orphans found.${NC}"
+    fi
+}
+
 # Pause after current session
 cmd_pause() {
     if [ ! -d "$LOKI_DIR" ]; then
@@ -4497,6 +4601,9 @@ main() {
         stop)
             cmd_stop
             ;;
+        cleanup)
+            cmd_cleanup "$@"
+            ;;
         pause)
             cmd_pause
             ;;

package/autonomy/run.sh

@@ -667,6 +667,146 @@ log_error() { echo -e "${RED}[ERROR]${NC} $*"; }
 log_step() { echo -e "${CYAN}[STEP]${NC} $*"; }
 log_debug() { [[ "${LOKI_DEBUG:-}" == "true" ]] && echo -e "${CYAN}[DEBUG]${NC} $*" || true; }
 
+#===============================================================================
+# Process Registry (PID Supervisor)
+# Central registry of all spawned child processes for reliable cleanup
+#===============================================================================
+
+PID_REGISTRY_DIR=""
+
+# Initialize the PID registry directory
+init_pid_registry() {
+    PID_REGISTRY_DIR="${TARGET_DIR:-.}/.loki/pids"
+    mkdir -p "$PID_REGISTRY_DIR"
+}
+
+# Parse a field from a JSON registry entry (python3 with shell fallback)
+# Usage: _parse_json_field <file> <field>
+_parse_json_field() {
+    local file="$1" field="$2"
+    if command -v python3 >/dev/null 2>&1; then
+        python3 -c "import json,sys; print(json.load(open(sys.argv[1])).get(sys.argv[2],''))" "$file" "$field" 2>/dev/null
+    else
+        # Shell fallback: extract value for simple flat JSON
+        sed 's/.*"'"$field"'":\s*//' "$file" 2>/dev/null | sed 's/[",}].*//' | head -1
+    fi
+}
+
+# Register a spawned process in the central registry
+# Usage: register_pid <pid> <label> [<extra_info>]
+# Example: register_pid $! "dashboard" "port=57374"
+register_pid() {
+    local pid="$1"
+    # Sanitize label and extra for JSON safety (escape backslash first, then double-quote, strip newlines)
+    local label="${2//\\/\\\\}"
+    label="${label//\"/\\\"}"
+    label="$(printf '%s' "$label" | tr -d '\n\r')"
+    local extra="${3:-}"
+    extra="${extra//\\/\\\\}"
+    extra="${extra//\"/\\\"}"
+    extra="$(printf '%s' "$extra" | tr -d '\n\r')"
+    [ -z "$PID_REGISTRY_DIR" ] && init_pid_registry
+    local entry_file="$PID_REGISTRY_DIR/${pid}.json"
+    cat > "$entry_file" << EOF
+{"pid":$pid,"label":"$label","started":"$(date -u +%Y-%m-%dT%H:%M:%SZ)","ppid":$$,"extra":"$extra"}
+EOF
+}
+
+# Unregister a process from the registry (called on clean shutdown)
+# Usage: unregister_pid <pid>
+unregister_pid() {
+    local pid="$1"
+    [ -z "$PID_REGISTRY_DIR" ] && init_pid_registry
+    rm -f "$PID_REGISTRY_DIR/${pid}.json" 2>/dev/null
+}
+
+# Kill a registered process with SIGTERM -> wait -> SIGKILL escalation
+# Usage: kill_registered_pid <pid>
+kill_registered_pid() {
+    local pid="$1"
+    if kill -0 "$pid" 2>/dev/null; then
+        kill "$pid" 2>/dev/null || true
+        # Wait up to 2 seconds for graceful exit
+        local waited=0
+        while [ $waited -lt 4 ] && kill -0 "$pid" 2>/dev/null; do
+            sleep 0.5
+            waited=$((waited + 1))
+        done
+        # Escalate to SIGKILL if still alive
+        if kill -0 "$pid" 2>/dev/null; then
+            kill -9 "$pid" 2>/dev/null || true
+        fi
+    fi
+    unregister_pid "$pid"
+}
+
+# Scan registry for orphaned processes and kill them
+# Called on startup and by `loki cleanup`
+# Returns: number of orphans killed
+cleanup_orphan_pids() {
+    [ -z "$PID_REGISTRY_DIR" ] && init_pid_registry
+    local orphan_count=0
+
+    if [ ! -d "$PID_REGISTRY_DIR" ]; then
+        echo "0"
+        return 0
+    fi
+
+    for entry_file in "$PID_REGISTRY_DIR"/*.json; do
+        [ -f "$entry_file" ] || continue
+        local pid
+        pid=$(basename "$entry_file" .json)
+
+        # Skip non-numeric filenames
+        case "$pid" in
+            ''|*[!0-9]*) continue ;;
+        esac
+
+        if kill -0 "$pid" 2>/dev/null; then
+            # Process is alive -- check if its parent session is dead
+            local ppid_val=""
+            ppid_val=$(_parse_json_field "$entry_file" "ppid") || true
+
+            # Validate ppid_val is numeric before using with kill
+            case "$ppid_val" in ''|*[!0-9]*) ppid_val="" ;; esac
+            if [ -n "$ppid_val" ] && [ "$ppid_val" != "$$" ]; then
+                if ! kill -0 "$ppid_val" 2>/dev/null; then
+                    # Parent is dead -- this is an orphan
+                    local label=""
+                    label=$(_parse_json_field "$entry_file" "label") || label="unknown"
+                    log_warn "Killing orphaned process: PID=$pid label=$label (parent $ppid_val is dead)" >&2
+                    kill_registered_pid "$pid"
+                    orphan_count=$((orphan_count + 1))
+                fi
+            fi
+        else
+            # Process is dead -- clean up stale registry entry
+            rm -f "$entry_file" 2>/dev/null
+        fi
+    done
+
+    echo "$orphan_count"
+}
+
+# Kill ALL registered processes (used during full shutdown)
+kill_all_registered() {
+    [ -z "$PID_REGISTRY_DIR" ] && init_pid_registry
+
+    if [ ! -d "$PID_REGISTRY_DIR" ]; then
+        return 0
+    fi
+
+    for entry_file in "$PID_REGISTRY_DIR"/*.json; do
+        [ -f "$entry_file" ] || continue
+        local pid
+        pid=$(basename "$entry_file" .json)
+        case "$pid" in
+            ''|*[!0-9]*) continue ;;
+        esac
+        kill_registered_pid "$pid"
+    done
+}
+
 #===============================================================================
 # Event Emission (Dashboard Integration)
 # Writes events to .loki/events.jsonl for dashboard consumption
@@ -1688,6 +1828,7 @@ create_worktree() {
         ) &
         # Capture install PID for cleanup on exit
         WORKTREE_INSTALL_PIDS+=($!)
+        register_pid "$!" "worktree-install" "stream=$stream_name"
 
         log_info "Created worktree: $worktree_path"
         return 0
@@ -1796,6 +1937,7 @@ spawn_worktree_session() {
 
     local pid=$!
     WORKTREE_PIDS[$stream_name]=$pid
+    register_pid "$pid" "worktree-session" "stream=$stream_name"
 
     log_info "Session spawned: $stream_name (PID: $pid)"
     return 0
@@ -2002,6 +2144,7 @@ cleanup_parallel_streams() {
         if kill -0 "$pid" 2>/dev/null; then
             kill "$pid" 2>/dev/null || true
         fi
+        unregister_pid "$pid"
     done
     WORKTREE_INSTALL_PIDS=()
 
@@ -2012,6 +2155,7 @@ cleanup_parallel_streams() {
             log_step "Stopping session: $stream"
             kill "$pid" 2>/dev/null || true
         fi
+        unregister_pid "$pid"
     done
 
     # Wait for all to finish
@@ -2620,8 +2764,8 @@ write_dashboard_state() {
     # Get complexity tier
     local complexity="${DETECTED_COMPLEXITY:-standard}"
 
-    # Get RARV cycle step (approximate based on iteration)
-    local rarv_step=$((ITERATION_COUNT % 4))
+    # Get RARV cycle step from actual phase tracking (falls back to iteration-based)
+    local rarv_step=${RARV_CURRENT_STEP:-$((ITERATION_COUNT % 4))}
     local rarv_stages='["reason", "act", "reflect", "verify"]'
 
     # Get memory system stats (if available)
@@ -2634,9 +2778,9 @@ write_dashboard_state() {
     [ -d ".loki/memory/skills" ] && procedural_count=$(find ".loki/memory/skills" -type f -name "*.json" 2>/dev/null | wc -l | tr -d ' ')
 
     # Get quality gates status (if available)
-    local quality_gates='{"staticAnalysis":"pending","codeReview":"pending","antiSycophancy":"pending","testCoverage":"pending","securityScan":"pending","performance":"pending"}'
+    local quality_gates='null'
     if [ -f ".loki/state/quality-gates.json" ]; then
-        quality_gates=$(cat ".loki/state/quality-gates.json" 2>/dev/null || echo "$quality_gates")
+        quality_gates=$(cat ".loki/state/quality-gates.json" 2>/dev/null || echo 'null')
     fi
 
     # Get Completion Council state (v5.25.0)
@@ -3037,6 +3181,7 @@ start_status_monitor() {
         done
     ) &
     STATUS_MONITOR_PID=$!
+    register_pid "$STATUS_MONITOR_PID" "status-monitor"
 
     log_info "Status monitor started"
     log_info "Monitor progress: ${CYAN}watch -n 2 cat .loki/STATUS.txt${NC}"
@@ -3046,6 +3191,7 @@ stop_status_monitor() {
     if [ -n "$STATUS_MONITOR_PID" ]; then
         kill "$STATUS_MONITOR_PID" 2>/dev/null || true
         wait "$STATUS_MONITOR_PID" 2>/dev/null || true
+        unregister_pid "$STATUS_MONITOR_PID"
     fi
     stop_resource_monitor
 }
@@ -3621,6 +3767,7 @@ start_resource_monitor() {
         done
     ) &
     RESOURCE_MONITOR_PID=$!
+    register_pid "$RESOURCE_MONITOR_PID" "resource-monitor"
 
     log_info "Resource monitor started (CPU threshold: ${RESOURCE_CPU_THRESHOLD}%, Memory threshold: ${RESOURCE_MEM_THRESHOLD}%)"
     log_info "Check status: ${CYAN}cat .loki/state/resources.json${NC}"
@@ -3630,6 +3777,7 @@ stop_resource_monitor() {
     if [ -n "$RESOURCE_MONITOR_PID" ]; then
         kill "$RESOURCE_MONITOR_PID" 2>/dev/null || true
         wait "$RESOURCE_MONITOR_PID" 2>/dev/null || true
+        unregister_pid "$RESOURCE_MONITOR_PID"
     fi
 }
 
@@ -4720,12 +4868,14 @@ BUILD_PROMPT
             esac
         ) &
         pids+=($!)
+        register_pid "$!" "code-reviewer" "name=$reviewer_name"
     done
 
     # Wait for all reviewers to complete
     log_info "Waiting for $reviewer_count reviewers to complete (blind review)..."
     for pid in "${pids[@]}"; do
         wait "$pid" || true
+        unregister_pid "$pid"
     done
 
     log_info "All reviewers complete. Aggregating verdicts..."
@@ -5191,6 +5341,7 @@ start_dashboard() {
     LOKI_TLS_CERT="${LOKI_TLS_CERT:-}" LOKI_TLS_KEY="${LOKI_TLS_KEY:-}" \
         LOKI_SKILL_DIR="${skill_dir}" PYTHONPATH="${skill_dir}" nohup "$python_cmd" -m dashboard.server > "$log_file" 2>&1 &
     DASHBOARD_PID=$!
+    register_pid "$DASHBOARD_PID" "dashboard" "port=${DASHBOARD_PORT:-57374}"
 
     # Save PID for later cleanup
     mkdir -p .loki/dashboard
@@ -5224,6 +5375,7 @@ stop_dashboard() {
     if [ -n "$DASHBOARD_PID" ]; then
         kill "$DASHBOARD_PID" 2>/dev/null || true
         wait "$DASHBOARD_PID" 2>/dev/null || true
+        unregister_pid "$DASHBOARD_PID"
     fi
 
     # Also try PID file
@@ -5231,6 +5383,7 @@ stop_dashboard() {
         local saved_pid=$(cat ".loki/dashboard/dashboard.pid" 2>/dev/null)
         if [ -n "$saved_pid" ]; then
             kill "$saved_pid" 2>/dev/null || true
+            unregister_pid "$saved_pid"
         fi
         rm -f ".loki/dashboard/dashboard.pid"
     fi
@@ -7121,6 +7274,7 @@ cleanup() {
         fi
         stop_dashboard
         stop_status_monitor
+        kill_all_registered
         rm -f "$loki_dir/loki.pid" 2>/dev/null
         if [ -f "$loki_dir/session.json" ]; then
             _LOKI_SESSION_FILE="$loki_dir/session.json" python3 -c "
@@ -7148,6 +7302,7 @@ except (json.JSONDecodeError, OSError): pass
         fi
         stop_dashboard
         stop_status_monitor
+        kill_all_registered
         rm -f .loki/loki.pid .loki/PAUSE 2>/dev/null
         # Mark session.json as stopped
         if [ -f ".loki/session.json" ]; then
@@ -7335,6 +7490,7 @@ main() {
         LOKI_RUNNING_FROM_TEMP='' nohup "$original_script" "${cmd_args[@]}" > "$log_file" 2>&1 &
         local bg_pid=$!
         echo "$bg_pid" > "$pid_file"
+        register_pid "$bg_pid" "background-session" "log=$log_file"
 
         echo ""
         echo -e "${GREEN}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${NC}"
@@ -7457,6 +7613,14 @@ main() {
     # Write PID file for ALL modes (foreground + background)
     echo "$$" > "$pid_file"
 
+    # Initialize PID registry and clean up orphans from previous sessions
+    init_pid_registry
+    local orphan_count
+    orphan_count=$(cleanup_orphan_pids)
+    if [ "$orphan_count" -gt 0 ]; then
+        log_warn "Killed $orphan_count orphaned process(es) from previous session"
+    fi
+
     # Copy skill files to .loki/skills/ - makes CLI self-contained
     # No need to install Claude Code skill separately
     copy_skill_files
@@ -7536,10 +7700,12 @@ main() {
             run_autonomous "$PRD_PATH"
         ) &
         local main_pid=$!
+        register_pid "$main_pid" "parallel-main" ""
 
         # Run parallel orchestrator
         run_parallel_orchestrator &
         local orchestrator_pid=$!
+        register_pid "$orchestrator_pid" "parallel-orchestrator" ""
 
         # Wait for main session (orchestrator continues watching)
         wait $main_pid || result=$?

package/dashboard/__init__.py

@@ -7,7 +7,7 @@ Modules:
     control: Session control API (start/stop/pause/resume)
 """
 
-__version__ = "5.49.0"
+__version__ = "5.49.1"
 
 # Expose the control app for easy import
 try:

package/dashboard/server.py

@@ -404,7 +404,20 @@ async def get_status() -> StatusResponse:
             iteration = state.get("iteration", 0)
             complexity = state.get("complexity", "standard")
             mode = state.get("mode", "")
-            running_agents = len(state.get("agents", []))
+            # Count only agents with alive PIDs (not raw array length)
+            agents_list = state.get("agents", [])
+            running_agents = 0
+            for agent in agents_list:
+                agent_pid = agent.get("pid")
+                if agent_pid:
+                    try:
+                        os.kill(int(agent_pid), 0)
+                        running_agents += 1
+                    except (OSError, ValueError, TypeError):
+                        pass
+                else:
+                    # No PID field -- count as running (legacy data)
+                    running_agents += 1
 
             tasks = state.get("tasks", {})
             pending_tasks = len(tasks.get("pending", []))
@@ -3194,6 +3207,31 @@ async def get_process_health(token: Optional[dict] = Depends(auth.get_current_to
         except Exception:
             pass
 
+    # PID registry (central process supervisor)
+    pids_dir = loki_dir / "pids"
+    registered: list[dict[str, Any]] = []
+    if pids_dir.exists():
+        for entry_file in sorted(pids_dir.glob("*.json")):
+            try:
+                pid_str = entry_file.stem
+                pid = int(pid_str)
+                entry = json.loads(entry_file.read_text())
+                try:
+                    os.kill(pid, 0)
+                    status = "alive"
+                except OSError:
+                    status = "dead"
+                registered.append({
+                    "pid": pid,
+                    "label": entry.get("label", "unknown"),
+                    "started": entry.get("started", ""),
+                    "ppid": entry.get("ppid"),
+                    "status": status,
+                })
+            except (ValueError, json.JSONDecodeError, OSError):
+                continue
+    result["registered_processes"] = registered
+
     watchdog_enabled = os.environ.get("LOKI_WATCHDOG", "false").lower() == "true"
     result["watchdog_enabled"] = watchdog_enabled
 

package/docs/COMPARISON.md

@@ -12,7 +12,7 @@
 | Feature | **Loki Mode** | **Zencoder** | **Devin** | **OpenAI Codex** | **Cursor** | **Claude Code** | **Kiro** | **Antigravity** | **Amazon Q** | **OpenCode** |
 |---------|--------------|--------------|-----------|-----------------|------------|-----------------|----------|-----------------|--------------|--------------|
 | **Type** | Skill/Framework | Enterprise Platform | Standalone Agent | Cloud Agent | AI IDE | CLI Agent | AI IDE | AI IDE | Cloud Agent | AI IDE (OSS) |
-| **Autonomy Level** | Full (zero human) | High | Full | High | Medium-High | High | High | High | High | High |
+| **Autonomy Level** | High (minimal human) | High | Full | High | Medium-High | High | High | High | High | High |
 | **Max Runtime** | Unlimited | Async/Scheduled | Hours | Per-task | Session | Session | Days | Async | Per-task | Session |
 | **Pricing** | Free (OSS) | Enterprise | $20/mo | ChatGPT Plus | $20/mo | API costs | Free preview | Free preview | $19/mo | Free (OSS) |
 | **Open Source** | Yes | No | No | No | No | No | No | No | No | Yes |
@@ -24,7 +24,7 @@
 
 | Feature | **Loki Mode** | **Devin** | **Codex** | **Cursor** | **Kiro** | **Antigravity** | **Amazon Q** | **OpenCode** |
 |---------|--------------|-----------|-----------|------------|----------|-----------------|--------------|--------------|
-| **Multi-Agent** | 41 agents in 7 swarms | Single | Single | Up to 8 parallel | Background | Manager Surface | Multiple types | 4 built-in |
+| **Multi-Agent** | 41 agents in 8 swarms | Single | Single | Up to 8 parallel | Background | Manager Surface | Multiple types | 4 built-in |
 | **Orchestration** | Full orchestrator | N/A | N/A | Git worktree | Hooks | Manager view | Workflow | Subagents |
 | **Parallel Exec** | 10+ Haiku, 4 impl (worktree) | No | No | 8 max | Yes | Yes | Yes | Yes |
 | **Agent Swarms** | Eng, Ops, Business, Data, Product, Growth, Review | N/A | N/A | N/A | N/A | N/A | 3 types | N/A |
@@ -180,9 +180,9 @@
 
 1. **Quality Control**: 7 explicit gates + blind review + devil's advocate vs built-in loops
 2. **Memory System**: 3-tier (episodic/semantic/procedural) with cross-project learning
-3. **Agent Specialization**: 41 pre-defined specialized agents across 7 swarms
+3. **Agent Specialization**: 41 pre-defined specialized agents across 8 swarms
 4. **Anti-Sycophancy**: CONSENSAGENT patterns prevent reviewer groupthink
-5. **Autonomy Design**: Zero human intervention from PRD to production
+5. **Autonomy Design**: Minimal human intervention from PRD to production
 6. **Research Foundation**: 10+ academic papers integrated vs proprietary
 
 ### Where Zencoder EXCEEDS Loki Mode
@@ -203,13 +203,13 @@
 |---------|--------------|---------|-----------------|------------|-----------------|---------------------|----------------|
 | **Stars** | 594 | 11,903 | 35K+ | 26K+ | 13.7K | N/A | N/A |
 | **npm/wk** | 6.1K | 21.4K | N/A | N/A | N/A | N/A | N/A |
-| **Agents** | 41 in 7 swarms | 11 agents | Fresh per task | 108 agents | Swarm-based | 32 agents | N/A |
+| **Agents** | 41 in 8 swarms | 11 agents | Fresh per task | 108 agents | Swarm-based | 32 agents | N/A |
 | **Skills** | Progressive disclosure | 6 slash commands | N/A | 129 skills | N/A | 35 skills | Memory focus |
 | **Multi-Provider** | Yes (Claude/Codex/Gemini) | 3 CLIs (separate) | No | No | No | No | No |
 | **Memory System** | 3-tier (episodic/semantic/procedural) | None | N/A | N/A | Hybrid | N/A | SQLite+FTS5 |
 | **Quality Gates** | 7 gates + Completion Council | User verify only | Two-Stage Review | N/A | Consensus | Tiered | N/A |
 | **Context Mgmt** | Standard | Fresh per task (core innovation) | Fresh per task | N/A | N/A | N/A | Progressive |
-| **Autonomy** | Full (zero human) | Semi (checkpoints) | Human-guided | Human-guided | Orchestrated | Human-guided | N/A |
+| **Autonomy** | High (minimal human) | Semi (checkpoints) | Human-guided | Human-guided | Orchestrated | Human-guided | N/A |
 
 ### What Loki Mode LACKS (Honest Assessment)
 
@@ -236,7 +236,7 @@ These are patterns from competing projects that are **practically and scientific
 | **Constitutional AI Integration** | Principles-based self-critique from Anthropic research | None have this |
 | **Anti-Sycophancy (CONSENSAGENT)** | Blind review + devil's advocate prevents groupthink | None have this |
 | **Provider Abstraction Layer** | Clean degradation from full-featured to sequential-only | Claude-only projects can't degrade |
-| **41 Specialized Agents** | Purpose-built agents in 7 swarms vs generic | agents (108) has more but less organized |
+| **41 Specialized Agents** | Purpose-built agents in 8 swarms vs generic | agents (108) has more but less organized |
 | **Research Foundation** | 10+ academic papers integrated with citations | Most have no research backing |
 
 ### Superpowers Deep-Dive (35K+ Stars)
@@ -342,7 +342,7 @@ Tiered agent architecture with explicit escalation:
 
 | Agent | Killer Feature |
 |-------|---------------|
-| **Loki Mode** | Zero-human-intervention full SDLC, 41 agents in 7 swarms, Constitutional AI, anti-sycophancy, cross-project learning, code transformation, property-based testing |
+| **Loki Mode** | Minimal-human-intervention full SDLC, 41 agents in 8 swarms, Constitutional AI, anti-sycophancy, cross-project learning, code transformation, property-based testing |
 | **Devin** | Full software engineer persona, Slack integration, 67% PR merge rate |
 | **OpenAI Codex** | Skills marketplace, $skill-creator, GPT-5.2-Codex, secure sandbox |
 | **Cursor** | 8 parallel agents, BugBot, Memories, $10B valuation, Composer model (250 tok/s) |
@@ -357,8 +357,8 @@ Tiered agent architecture with explicit escalation:
 
 | Dimension | Loki Mode Advantage |
 |-----------|-------------------|
-| **Autonomy** | Only agent designed for TRUE zero human intervention |
-| **Multi-Agent** | 41 specialized agents in 7 swarms vs 1-8 in competitors |
+| **Autonomy** | Designed for high autonomy with minimal human intervention |
+| **Multi-Agent** | 41 specialized agents in 8 swarms vs 1-8 in competitors |
 | **Quality** | 7 gates + blind review + devil's advocate + property-based testing |
 | **Research** | 10+ academic papers integrated vs proprietary/undisclosed |
 | **Anti-Sycophancy** | Only agent with CONSENSAGENT-based blind review |

package/docs/COMPETITIVE-ANALYSIS.md

@@ -37,9 +37,9 @@ GSD is the closest competitor -- a context engineering system that spawns fresh
 |---------|-----------|-------------|---------|--------|--------------|-------|
 | **GitHub Stars** | 594 | 13,700 | 62,400 | 25,000+ | N/A (Commercial) | N/A (Commercial) |
 | **Agent Count** | 41 types | 64+ agents | 5 roles | Unlimited | 8 parallel | 1 autonomous |
-| **Parallel Execution** | Yes (100+) | Yes (swarms) | Sequential | Yes (crews) | Yes (8 worktrees) | Yes (fleet) |
-| **Published Benchmarks** | **98.78% HumanEval (multi-agent)** | None | 85.9-87.7% HumanEval | None | ~250 tok/s | 15% complex tasks |
-| **SWE-bench Score** | **99.67% patch gen (299/300)** | Unknown | Unknown | Unknown | Unknown | 15% complex |
+| **Parallel Execution** | Yes (multi-agent) | Yes (swarms) | Sequential | Yes (crews) | Yes (8 worktrees) | Yes (fleet) |
+| **Published Benchmarks** | 98.78% HumanEval (self-reported, max 3 retries) | None | 85.9-87.7% HumanEval | None | ~250 tok/s | 15% complex tasks |
+| **SWE-bench Score** | 99.67% patch gen (unevaluated, 299/300) | Unknown | Unknown | Unknown | Unknown | 15% complex |
 | **Full SDLC** | Yes (8 phases) | Yes | Partial | Partial | No | Partial |
 | **Business Ops** | **Yes (8 agents)** | No | No | No | No | No |
 | **Enterprise Security** | `--dangerously-skip-permissions` | MCP sandboxed | Sandboxed | Audit logs, RBAC | Staged autonomy | Sandboxed |

package/docs/INSTALLATION.md

@@ -2,11 +2,11 @@
 
 The flagship product of [Autonomi](https://www.autonomi.dev/). Complete installation instructions for all platforms and use cases.
 
-**Version:** v5.49.0
+**Version:** v5.49.1
 
 ---
 
-## What's New in v5.39.0
+## What's New in v5.49.1
 
 ### Enterprise Security (v5.36.0-v5.37.1)
 - TLS/HTTPS support for dashboard connections
@@ -63,7 +63,7 @@ npm install -g loki-mode
 brew tap asklokesh/tap && brew install loki-mode
 
 # Option C: Docker
-docker pull asklokesh/loki-mode:5.32.0
+docker pull asklokesh/loki-mode:latest
 
 # Option D: Git clone
 git clone https://github.com/asklokesh/loki-mode.git ~/.claude/skills/loki-mode
@@ -160,6 +160,10 @@ Install via npm for the easiest setup with automatic PATH configuration.
 npm install -g loki-mode
 
 # The skill is automatically installed to ~/.claude/skills/loki-mode
+
+# Opt out of anonymous install telemetry:
+# LOKI_TELEMETRY_DISABLED=true npm install -g loki-mode
+# Or set DO_NOT_TRACK=1
 ```
 
 ### Usage
@@ -207,8 +211,8 @@ brew tap asklokesh/tap
 # Install Loki Mode
 brew install loki-mode
 
-# Set up Claude Code skill integration
-loki-mode-install-skill
+# Set up Claude Code skill integration (manual symlink required)
+ln -sf "$(brew --prefix)/opt/loki-mode/libexec" ~/.claude/skills/loki-mode
 ```
 
 ### Dependencies
@@ -254,7 +258,7 @@ Run Loki Mode in a container for isolated execution.
 
 ```bash
 # Pull the image
-docker pull asklokesh/loki-mode:5.32.0
+docker pull asklokesh/loki-mode:latest
 
 # Or use docker-compose
 curl -o docker-compose.yml https://raw.githubusercontent.com/asklokesh/loki-mode/main/docker-compose.yml
@@ -264,10 +268,10 @@ curl -o docker-compose.yml https://raw.githubusercontent.com/asklokesh/loki-mode
 
 ```bash
 # Run with a PRD file
-docker run -v $(pwd):/workspace -w /workspace asklokesh/loki-mode:5.32.0 start ./my-prd.md
+docker run -v $(pwd):/workspace -w /workspace asklokesh/loki-mode:latest start ./my-prd.md
 
 # Interactive mode
-docker run -it -v $(pwd):/workspace -w /workspace asklokesh/loki-mode:5.32.0
+docker run -it -v $(pwd):/workspace -w /workspace asklokesh/loki-mode:latest
 
 # Using docker-compose
 docker-compose run loki start ./my-prd.md
@@ -280,7 +284,7 @@ Pass your configuration via environment variables:
 ```bash
 docker run -e LOKI_MAX_RETRIES=100 -e LOKI_BASE_WAIT=120 \
   -v $(pwd):/workspace -w /workspace \
-  asklokesh/loki-mode:5.32.0 start ./my-prd.md
+  asklokesh/loki-mode:latest start ./my-prd.md
 ```
 
 ### Updating
@@ -396,12 +400,12 @@ Pass the provider as an environment variable:
 # Use Codex with Docker
 docker run -e LOKI_PROVIDER=codex \
   -v $(pwd):/workspace -w /workspace \
-  asklokesh/loki-mode:5.32.0 start ./my-prd.md
+  asklokesh/loki-mode:latest start ./my-prd.md
 
 # Use Gemini with Docker
 docker run -e LOKI_PROVIDER=gemini \
   -v $(pwd):/workspace -w /workspace \
-  asklokesh/loki-mode:5.32.0 start ./my-prd.md
+  asklokesh/loki-mode:latest start ./my-prd.md
 ```
 
 ### Degraded Mode
@@ -652,7 +656,11 @@ Add the source command to your startup file so completions load every time you o
 Add this line to your `~/.bashrc` (Linux) or `~/.bash_profile` (macOS):
 
 ```bash
-source /path/to/loki/completions/loki.bash
+# npm install: use the npm package path
+source "$(npm root -g)/loki-mode/completions/loki.bash"
+
+# git clone: use the skills directory
+source ~/.claude/skills/loki-mode/completions/loki.bash
 ```
 
 ---

package/docs/auto-claude-comparison.md

@@ -169,7 +169,7 @@ Loki Mode:
 **Verdict: Loki Mode wins** - Simpler, lighter footprint.
 
 ### 10. Cursor Scale Patterns (v3.3.0)
-Loki Mode now incorporates proven patterns from Cursor's 100+ agent deployments:
+Loki Mode now incorporates proven patterns from Cursor's large-scale agent deployments:
 - Recursive sub-planners
 - Judge agents for cycle decisions
 - Optimistic concurrency control

package/docs/cursor-comparison.md

@@ -9,7 +9,7 @@
 
 | Dimension | Cursor | Loki Mode | Winner |
 |-----------|--------|-----------|--------|
-| **Proven Scale** | 1M+ LoC, 100+ agents | Benchmarks only | Cursor |
+| **Proven Scale** | 1M+ LoC, large agent count | Benchmarks only | Cursor |
 | **Research Foundation** | Empirical iteration | 25+ academic citations | Loki Mode |
 | **Quality Assurance** | Workers self-manage | 7-gate system + anti-sycophancy | Loki Mode |
 | **Anti-Sycophancy** | Not mentioned | CONSENSAGENT blind review | Loki Mode |
@@ -122,7 +122,7 @@ BOOTSTRAP -> DISCOVERY -> ARCHITECTURE -> INFRASTRUCTURE
      -> DEVELOPMENT -> QA -> DEPLOYMENT -> GROWTH (continuous)
 ```
 
-**41 Specialized Agent Types across 7 swarms:**
+**41 Specialized Agent Types across 8 swarms:**
 - Engineering (8 types)
 - Operations (8 types)
 - Business (8 types)
@@ -184,7 +184,7 @@ We incorporated Cursor's proven patterns:
 
 1. **Recursive Sub-Planners** - Planning scales horizontally
 2. **Judge Agents** - Explicit CONTINUE/COMPLETE/ESCALATE/PIVOT decisions
-3. **Optimistic Concurrency** - No locks, scales to 100+ agents
+3. **Optimistic Concurrency** - No locks, scales horizontally
 4. **Scale-Aware Review** - Full review for high-risk only at scale
 
 ---

package/docs/thick2thin.md

@@ -49,7 +49,7 @@ skills/
 
 references/ (unchanged)
   +-- 18 detailed reference files
-  +-- agents.md (23KB) - Full 37 agent specs
+  +-- agents.md (23KB) - Full 41 agent specs
   +-- openai-patterns.md, lab-research-patterns.md, etc.
 ```
 
@@ -117,7 +117,7 @@ Content that didn't exist in v2.38.0:
 | Handoff message format | A2A specification | skills/agents.md |
 | Agentic patterns table | awesome-agentic-patterns | skills/agents.md |
 | "Ralph Wiggum Mode" insight | moridinamael | skills/agents.md |
-| Full 37 agent reference | references/agent-types.md | skills/agents.md (pointer) |
+| Full 41 agent reference | references/agent-types.md | skills/agents.md (pointer) |
 | References directory listing | New | skills/00-index.md |
 
 ---

package/mcp/__init__.py

@@ -21,4 +21,4 @@ try:
 except ImportError:
     __all__ = ['mcp']
 
-__version__ = '5.49.0'
+__version__ = '5.49.1'

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "loki-mode",
-  "version": "5.49.0",
+  "version": "5.49.1",
   "description": "Loki Mode by Autonomi - Multi-agent autonomous startup system for Claude Code, Codex CLI, and Gemini CLI",
   "keywords": [
     "autonomi",

package/references/agent-types.md

@@ -6,7 +6,7 @@ Complete definitions and capabilities for all 41 specialized agent types.
 
 ## Overview
 
-Loki Mode has 41 predefined agent types organized into 7 specialized swarms (37 domain agents + 4 orchestration agents). The orchestrator spawns only the agents needed for your project - a simple app might use 5-10 agents, while a complex startup could spawn 100+ agents working in parallel.
+Loki Mode has 41 predefined agent types organized into 8 specialized swarms (37 domain agents + 4 orchestration agents). The orchestrator spawns only the agents needed for your project -- typically 5-10 for simple projects, more for complex ones.
 
 ---
 
@@ -98,7 +98,7 @@ Loki Mode has 41 predefined agent types organized into 7 specialized swarms (37
 
 ## Orchestration Swarm (4 types)
 
-> **Source:** [Cursor Scaling Learnings](./cursor-learnings.md) - patterns proven at 100+ agent scale
+> **Source:** [Cursor Scaling Learnings](./cursor-learnings.md) - patterns proven at large agent scale
 
 | Agent | Capabilities |
 |-------|-------------|

package/references/agents.md

@@ -2,7 +2,7 @@
 
 Complete specifications for all 41 specialized agent types in the Loki Mode multi-agent system (37 domain agents + 4 orchestration agents).
 
-**Note:** These are agent TYPE definitions, not a fixed count. Loki Mode dynamically spawns agents based on project needs - a simple todo app might use 5-10 agents, while a complex startup could spawn 100+ agents working in parallel.
+**Note:** These are agent TYPE definitions, not a fixed count. Loki Mode dynamically spawns agents based on project needs - a simple todo app might use 5-10 agents, while a complex startup spawns more as needed.
 
 ## Agent Role Prompt Template
 

package/references/competitive-analysis.md

@@ -182,7 +182,7 @@ Dexter shows value of domain specialization. Our 41 agent types follow this patt
    - Most haven't scaled across enterprise
 
 ### Loki Mode Alignment
-- Multi-agent architecture (41 types, 7 swarms)
+- Multi-agent architecture (41 types, 8 swarms)
 - Plan Agents (orchestrator, planner)
 - Execution Agents (eng-*, ops-*, biz-*)
 - Security controls (LOKI_SANDBOX_MODE, LOKI_BLOCKED_COMMANDS)

package/skills/agents.md

@@ -1,6 +1,6 @@
 # Agent Dispatch & Structured Prompting
 
-> **Full agent type definitions:** See `references/agent-types.md` for complete 41 agent role specifications across 7 swarms (Engineering, Operations, Business, Data, Product, Growth, Review, Orchestration).
+> **Full agent type definitions:** See `references/agent-types.md` for complete 41 agent role specifications across 8 swarms (Engineering, Operations, Business, Data, Product, Growth, Review, Orchestration).
 
 ---
 
@@ -245,7 +245,7 @@ Priority order for context:
 
 ---
 
-## The 37 Agent Roles
+## The 41 Agent Roles (37 Domain + 4 Orchestration)
 
 See `references/agent-types.md` for complete specifications. Summary:
 
@@ -259,4 +259,4 @@ See `references/agent-types.md` for complete specifications. Summary:
 | Growth | hacker, community, success, lifecycle | 4 |
 | Review | code, business, security | 3 |
 
-**Spawn only what you need.** Simple project: 5-10 agents. Complex startup: 100+.
+**Spawn only what you need.** Simple project: 5-10 agents. Complex startup: more as needed.

package/skills/parallel-workflows.md

@@ -427,7 +427,7 @@ optimistic_write:
     - No waiting for locks
     - No deadlock risk
     - Failed writes are cheap (just retry)
-    - Scales to 100+ agents
+    - Scales horizontally with agent count
 ```
 
 ### Implementation

package/skills/quality-gates.md

@@ -472,7 +472,7 @@ See `references/quality-control.md` for complete details.
 
 ## Scale Considerations
 
-> **Source:** [Cursor Scaling Learnings](../references/cursor-learnings.md) - integrators became bottlenecks at 100+ agents
+> **Source:** [Cursor Scaling Learnings](../references/cursor-learnings.md) - integrators became bottlenecks at high agent counts
 
 ### Review Intensity Scaling