livepilot 1.9.14 → 1.9.16

package/livepilot/skills/livepilot-core/SKILL.md

@@ -1,17 +1,17 @@
 ---
 name: livepilot-core
-description: Core discipline for LivePilot — agentic production system for Ableton Live 12. 178 tools across 17 domains. Device atlas (280+ devices), M4L analyzer (spectrum/RMS/key detection), technique memory, automation intelligence (16 curve types, 15 recipes), music theory (Krumhansl-Schmuckler, species counterpoint), generative algorithms (Euclidean rhythm, tintinnabuli, phase shift), neo-Riemannian harmony (PRL transforms, Tonnetz), MIDI file I/O. Use whenever working with Ableton Live through MCP tools.
+description: Core discipline for LivePilot — agentic production system for Ableton Live 12. 236 tools across 32 domains. Device atlas (280+ devices), M4L analyzer (spectrum/RMS/key detection), technique memory, automation intelligence (16 curve types, 15 recipes), music theory (Krumhansl-Schmuckler, species counterpoint), generative algorithms (Euclidean rhythm, tintinnabuli, phase shift), neo-Riemannian harmony (PRL transforms, Tonnetz), MIDI file I/O. Use whenever working with Ableton Live through MCP tools.
 ---
 
 # LivePilot Core — Ableton Live 12
 
-Agentic production system for Ableton Live 12. 178 tools across 17 domains, three layers:
+Agentic production system for Ableton Live 12. 236 tools across 32 domains, three layers:
 
 - **Device Atlas** — A structured knowledge corpus of 280+ instruments, 139 drum kits, and 350+ impulse responses. Consult the atlas before loading any device. It contains real browser URIs, preset names, and sonic descriptions. Never guess a device name — look it up.
 - **M4L Analyzer** — Real-time audio analysis on the master bus (8-band spectrum, RMS/peak, key detection). Use it to verify mixing decisions, detect frequency problems, and find the key before writing harmonic content.
 - **Technique Memory** — Persistent storage for production decisions. Consult `memory_recall` before creative tasks to understand the user's taste. Save techniques when the user likes something. The memory shapes future decisions without constraining them.
 
-These layers sit on top of 178 deterministic tools across 17 domains: transport, tracks, clips, notes, devices, scenes, mixing, browser, arrangement, memory, analyzer, automation, theory, generative, harmony, MIDI I/O, and perception.
+These layers sit on top of 236 deterministic tools across 32 domains: transport, tracks, clips, notes, devices, scenes, mixing, browser, arrangement, memory, analyzer, automation, theory, generative, harmony, MIDI I/O, perception, agent_os, composition, motif, research, planner, project_brain, runtime, evaluation, memory_fabric, mix_engine, sound_design, transition_engine, reference_engine, translation_engine, and performance_engine.
 
 ## Golden Rules
 
@@ -35,7 +35,7 @@ These layers sit on top of 178 deterministic tools across 17 domains: transport,
 Not all tools respond instantly. Know the tiers and act accordingly.
 
 ### Instant (<1s) — Use freely, no warning needed
-All 178 core tools (transport, tracks, clips, notes, devices, scenes, mixing, browser, arrangement, memory, automation, theory, generative, harmony, midi_io, perception) plus Layer A perception tools (spectral shape, timbral profile, mel spectrum, chroma, onsets, harmonic/percussive, novelty, momentary loudness). These are the reflex tools — call them anytime without hesitation.
+All 236 core tools (transport, tracks, clips, notes, devices, scenes, mixing, browser, arrangement, memory, automation, theory, generative, harmony, midi_io, perception) plus Layer A perception tools (spectral shape, timbral profile, mel spectrum, chroma, onsets, harmonic/percussive, novelty, momentary loudness). These are the reflex tools — call them anytime without hesitation.
 
 ### Fast (1-5s) — Use freely, barely noticeable
 `analyze_loudness` · `analyze_dynamic_range` · `compare_loudness`
@@ -133,7 +133,7 @@ Never skip levels. The user's question determines the entry point, but always st
 - **Dead AU/VST plugin** — `parameter_count` <= 1 on a PluginDevice (plugin shell loaded, DSP engine crashed)
 - **Sample-dependent plugin with no sample** — granular synths, bare samplers, and sample players load "successfully" with many parameters but produce zero audio without source material. The sneakiest silent failure because `get_device_info` looks healthy.
 
-## Tool Domains (178 total)
+## Tool Domains (236 total)
 
 ### Transport (12)
 `get_session_info` · `set_tempo` · `set_time_signature` · `start_playback` · `stop_playback` · `continue_playback` · `toggle_metronome` · `set_session_loop` · `undo` · `redo` · `get_recent_actions` · `get_session_diagnostics`
@@ -249,6 +249,81 @@ MIDI file import/export — works with standard .mid files on disk.
 - `extract_piano_roll` returns a 2D velocity matrix (pitch × time) from a .mid file for visualization or processing
 - Dependencies: midiutil (export), pretty-midi (import/analysis) — lazy-loaded, ~5 MB total
 
+### Agent OS (8)
+Goal-driven decision loop — compile goals, build world models, evaluate moves, learn from outcomes.
+
+**Tools:** `compile_goal_vector` · `build_world_model` · `evaluate_move` · `analyze_outcomes` · `get_technique_card` · `get_taste_profile` · `get_turn_budget` · `route_request`
+
+### Composition (9)
+Large-scale arrangement structure — sections, phrases, gestures, harmonic fields, transitions.
+
+**Tools:** `analyze_composition` · `get_section_graph` · `get_phrase_grid` · `plan_gesture` · `evaluate_composition_move` · `get_harmony_field` · `get_transition_analysis` · `apply_gesture_template` · `get_section_outcomes`
+
+### Motif (2)
+Recurring pattern detection and classical transformation.
+
+**Tools:** `get_motif_graph` · `transform_motif`
+
+### Research (3)
+Production technique lookup, emotional arc analysis, and genre-specific tactics.
+
+**Tools:** `research_technique` · `get_emotional_arc` · `get_style_tactics`
+
+### Planner (2)
+Arrangement planning — transform loops into full structures.
+
+**Tools:** `plan_arrangement` · `transform_section`
+
+### Project Brain (2)
+Comprehensive project model — tracks, sections, capabilities, staleness.
+
+**Tools:** `build_project_brain` · `get_project_brain_summary`
+
+### Runtime (4)
+Capability state, action ledger, and safety validation.
+
+**Tools:** `get_capability_state` · `get_action_ledger_summary` · `get_last_move` · `check_safety`
+
+### Evaluation (1)
+Unified move evaluation using the Evaluation Fabric.
+
+**Tools:** `evaluate_with_fabric`
+
+### Memory Fabric (6)
+Anti-preferences, session memory, taste dimensions, and promotion candidates.
+
+**Tools:** `get_anti_preferences` · `record_anti_preference` · `get_promotion_candidates` · `get_session_memory` · `add_session_memory` · `get_taste_dimensions`
+
+### Mix Engine (6)
+Spectral mix analysis, issue detection, move planning, and evaluation.
+
+**Tools:** `analyze_mix` · `get_mix_issues` · `plan_mix_move` · `evaluate_mix_move` · `get_masking_report` · `get_mix_summary`
+
+### Sound Design (4)
+Device chain analysis, issue detection, and move planning per track.
+
+**Tools:** `analyze_sound_design` · `get_sound_design_issues` · `plan_sound_design_move` · `get_patch_model`
+
+### Transition Engine (3)
+Section transition analysis, planning, and scoring.
+
+**Tools:** `analyze_transition` · `plan_transition` · `score_transition`
+
+### Reference Engine (3)
+Reference profile building, gap analysis, and move planning.
+
+**Tools:** `build_reference_profile` · `analyze_reference_gaps` · `plan_reference_moves`
+
+### Translation Engine (2)
+Playback robustness — mono safety, small speakers, harshness detection.
+
+**Tools:** `check_translation` · `get_translation_issues`
+
+### Performance Engine (3)
+Live performance support — scene state, safe moves, and handoffs.
+
+**Tools:** `get_performance_state` · `get_performance_safe_moves` · `plan_scene_handoff`
+
 ## Workflow: Building a Beat
 
 1. `get_session_info` — check current state
@@ -397,7 +472,7 @@ Deep production knowledge lives in `references/`. Consult these when making crea
 
 | File | What's inside | When to consult |
 |------|--------------|-----------------|
-| `references/overview.md` | All 178 tools mapped with params, units, ranges | Quick lookup for any tool |
+| `references/overview.md` | All 236 tools mapped with params, units, ranges | Quick lookup for any tool |
 | `references/midi-recipes.md` | Drum patterns by genre, chord voicings, scales, hi-hat techniques, humanization, polymetrics | Programming MIDI notes, building beats |
 | `references/sound-design.md` | Stock instruments/effects, parameter recipes for bass/pad/lead/pluck, device chain patterns | Loading and configuring devices |
 | `references/mixing-patterns.md` | Gain staging, parallel compression, sidechain, EQ by instrument, bus processing, stereo width | Setting volumes, panning, adding effects |

package/livepilot/skills/livepilot-core/references/m4l-devices.md

@@ -32,7 +32,7 @@ Ableton's browser is the source for all devices, presets, and samples. The `sear
 | **Delay** | Delay, Echo, Grain Delay, Beat Repeat, Spectral Time |
 | **Reverb** | Reverb, Convolution Reverb (M4L), Hybrid Reverb |
 | **Distortion** | Saturator, Overdrive, Erosion, Redux, Pedal, Amp, Cabinet |
-| **Modulation** | Chorus-Ensemble, Phaser-Flanger, Frequency Shifter, Ring Mod |
+| **Modulation** | Chorus-Ensemble, Phaser-Flanger, Shifter, Ring Mod |
 | **Utility** | Utility, Tuner, Spectrum, External Audio Effect |
 | **Spatial** | Surround Panner (if available) |
 
@@ -295,7 +295,7 @@ find_and_load_device(track_index=0, name="Arpeggiator")
 | Delay | `Delay`, `Echo`, `Grain Delay`, `Beat Repeat` |
 | Reverb | `Reverb`, `Hybrid Reverb` |
 | Distortion | `Saturator`, `Overdrive`, `Erosion`, `Redux`, `Pedal`, `Amp`, `Cabinet` |
-| Modulation | `Chorus-Ensemble`, `Phaser-Flanger`, `Frequency Shifter` |
+| Modulation | `Chorus-Ensemble`, `Phaser-Flanger`, `Shifter` |
 | Utility | `Utility`, `Tuner`, `Spectrum` |
 | MIDI FX | `Arpeggiator`, `Chord`, `Note Length`, `Pitch`, `Random`, `Scale`, `Velocity`, `CC Control` |
 | Racks | `Instrument Rack`, `Audio Effect Rack`, `MIDI Effect Rack` |

package/livepilot/skills/livepilot-core/references/overview.md

@@ -1,6 +1,6 @@
-# LivePilot v1.9.14 — Architecture & Tool Reference
+# LivePilot v1.9.16 — Architecture & Tool Reference
 
-Agentic production system for Ableton Live 12. 178 tools across 17 domains. Device atlas (280+ devices), spectral perception (M4L analyzer), technique memory, automation intelligence (16 curve types, 15 recipes), music theory (Krumhansl-Schmuckler, species counterpoint), generative algorithms (Euclidean rhythm, tintinnabuli, phase shift, additive process), neo-Riemannian harmony (PRL transforms, Tonnetz), MIDI file I/O.
+Agentic production system for Ableton Live 12. 236 tools across 32 domains. Device atlas (280+ devices), spectral perception (M4L analyzer), technique memory, automation intelligence (16 curve types, 15 recipes), music theory (Krumhansl-Schmuckler, species counterpoint), generative algorithms (Euclidean rhythm, tintinnabuli, phase shift, additive process), neo-Riemannian harmony (PRL transforms, Tonnetz), MIDI file I/O.
 
 ## Architecture
 
@@ -32,7 +32,7 @@ A flat tool list lets the AI press buttons. LivePilot's three layers give it con
 
 This turns "set EQ band 3 to -4 dB" into "cut 400 Hz by 4 dB, then read the spectrum to confirm the mud is actually reduced."
 
-## The 178 Tools — What Each One Does
+## The 236 Tools — What Each One Does
 
 ### Transport (12) — Playback, tempo, global state, diagnostics
 

package/livepilot/skills/livepilot-core/references/sound-design.md

@@ -215,11 +215,12 @@ Simple 3-band EQ for quick tonal shaping.
 - `Feedback` — Intensity/resonance
 - **Phaser**: Sweeping notches. **Flanger**: Jet/whoosh effect.
 
-**Frequency Shifter**
+**Shifter** (renamed from Frequency Shifter in Live 12)
 - `Frequency` — Shift amount in Hz (not semitones — inharmonic)
 - `Drive` — Input gain
 - `Dry/Wet` — Mix
 - **Use for**: Metallic textures, detuned unease, ring-mod effects
+- **Browser name**: `Shifter` (use this with `find_and_load_device`)
 
 ### Utility
 
@@ -320,7 +321,7 @@ LFO: Multiple slow LFOs mapped to position, filter, pitch (subtle)
 ```
 - Heavy Reverb (5-10s decay, high diffusion, 80-100% wet)
 - Delay with high feedback (60-80%), filtered
-- Frequency Shifter at very small values (+/- 1-5 Hz) for movement
+- Shifter at very small values (+/- 1-5 Hz) for movement
 - Grain Delay for granular textures
 
 ### 808 Bass (Simpler or Drum Rack)

package/livepilot/skills/livepilot-release/SKILL.md

@@ -28,26 +28,26 @@ Run this checklist EVERY time the user says "update everything", "push", "releas
 
 ## 2. Tool Count (must ALL match)
 
-Current: **178 tools across 17 domains**.
+Current: **236 tools across 32 domains**.
 Core (no M4L): **149**. Analyzer (M4L): **29**. Perception (offline): **4**.
 
 Verify: `grep -rc "@mcp.tool" mcp_server/tools/ | grep -v ":0" | awk -F: '{sum+=$2} END{print sum}'`
 
 Files that reference tool count:
-- [ ] `README.md` — header, PERCEPTION section ("149 core...29 analyzer"), Analyzer table header "(29)", Perception table header "(4)"
-- [ ] `package.json` → `"description"` (178 tools, 17 domains)
+- [ ] `README.md` — header, PERCEPTION section ("207 core...29 analyzer"), Analyzer table header "(29)", Perception table header "(4)"
+- [ ] `package.json` → `"description"` (236 tools, 32 domains)
 - [ ] `server.json` → `"description"`
 - [ ] `livepilot/.Codex-plugin/plugin.json` → `"description"` (primary Codex manifest)
 - [ ] `livepilot/.claude-plugin/plugin.json` → `"description"` (must match Codex plugin)
 - [ ] `.claude-plugin/marketplace.json` → `"description"`
-- [ ] `CLAUDE.md` → "178 tools across 17 domains"
-- [ ] `livepilot/skills/livepilot-core/SKILL.md` — "178 tools across 17 domains", Analyzer (29), Perception (4)
-- [ ] `livepilot/skills/livepilot-core/references/overview.md` — "178 tools across 17 domains"
+- [ ] `CLAUDE.md` → "236 tools across 32 domains"
+- [ ] `livepilot/skills/livepilot-core/SKILL.md` — "236 tools across 32 domains", Analyzer (29), Perception (4)
+- [ ] `livepilot/skills/livepilot-core/references/overview.md` — "236 tools across 32 domains"
 - [ ] `docs/manual/index.md` — domain table: Analyzer (29), Perception (4)
-- [ ] `docs/manual/getting-started.md` — "149 core tools...29 analyzer"
+- [ ] `docs/manual/getting-started.md` — "207 core tools...29 analyzer"
 - [ ] `docs/manual/tool-reference.md` — all domains present with correct counts
 - [ ] `docs/TOOL_REFERENCE.md` — all domains present
-- [ ] `docs/M4L_BRIDGE.md` — "149 core tools...29 analyzer"
+- [ ] `docs/M4L_BRIDGE.md` — "207 core tools...29 analyzer"
 - [ ] `docs/social-banner.html`
 - [ ] `mcp_server/tools/analyzer.py` → module docstring
 - [ ] `tests/test_tools_contract.py` → expected total count
@@ -56,10 +56,10 @@ Files that reference tool count:
 
 ## 3. Domain Count
 
-Current: **17 domains**: transport, tracks, clips, notes, devices, scenes, mixing, browser, arrangement, memory, analyzer, automation, theory, generative, harmony, midi_io, perception.
+Current: **32 domains**: transport, tracks, clips, notes, devices, scenes, mixing, browser, arrangement, memory, analyzer, automation, theory, generative, harmony, midi_io, perception, agent_os, composition, research, planner, project_brain, runtime, evaluation, memory_fabric, mix_engine, sound_design, transition_engine, reference_engine, translation_engine, performance_engine.
 
-- [ ] All files that mention domain count say "17 domains"
-- [ ] Domain lists include ALL 17 (especially perception — it's the newest and most often omitted)
+- [ ] All files that mention domain count say "32 domains"
+- [ ] Domain lists include ALL 32 (especially newer domains — they're the most often omitted)
 
 ## 4. npm Registry
 
@@ -89,8 +89,8 @@ Current: **17 domains**: transport, tracks, clips, notes, devices, scenes, mixin
 
 - [ ] `README.md` — features match current capabilities, "Coming" section is accurate
 - [ ] `docs/manual/getting-started.md` — install instructions current
-- [ ] `docs/manual/tool-reference.md` — all 17 domains listed, all 178 tools present
-- [ ] `docs/TOOL_REFERENCE.md` — all 17 domains present
+- [ ] `docs/manual/tool-reference.md` — all 32 domains listed, all 236 tools present
+- [ ] `docs/TOOL_REFERENCE.md` — all 32 domains present
 - [ ] `docs/M4L_BRIDGE.md` — architecture accurate, core tool count correct
 
 ## 9. Derived Artifacts

package/m4l_device/livepilot_bridge.js

@@ -84,7 +84,7 @@ function anything() {
 function dispatch(cmd, args) {
     switch(cmd) {
         case "ping":
-            send_response({"ok": true, "version": "1.9.14"});
+            send_response({"ok": true, "version": "1.9.16"});
             break;
         case "get_params":
             cmd_get_params(args);
@@ -513,16 +513,17 @@ function cmd_get_selected() {
         appointed_device: null
     };
 
-    // Selected track
+    // Selected track — match by object ID (not name, which can be duplicated)
     try {
         cursor_b.goto("live_set view selected_track");
         result.selected_track_name = cursor_b.get("name").toString();
-        // Get track index by walking tracks
+        var selected_id = cursor_b.id;
+        // Get track index by walking tracks and comparing IDs
         cursor_a.goto("live_set");
         var tc = cursor_a.getcount("tracks");
         for (var i = 0; i < tc; i++) {
             cursor_a.goto("live_set tracks " + i);
-            if (cursor_a.get("name").toString() === result.selected_track_name) {
+            if (cursor_a.id === selected_id) {
                 result.selected_track = i;
                 break;
             }
@@ -533,7 +534,7 @@ function cmd_get_selected() {
             var rtc = cursor_a.getcount("return_tracks");
             for (var j = 0; j < rtc; j++) {
                 cursor_a.goto("live_set return_tracks " + j);
-                if (cursor_a.get("name").toString() === result.selected_track_name) {
+                if (cursor_a.id === selected_id) {
                     result.selected_track = -(j + 1);  // -1, -2, ... convention
                     break;
                 }
@@ -542,7 +543,7 @@ function cmd_get_selected() {
         // Check master track if still not found
         if (result.selected_track === -1) {
             cursor_a.goto("live_set master_track");
-            if (cursor_a.get("name").toString() === result.selected_track_name) {
+            if (cursor_a.id === selected_id) {
                 result.selected_track = -1000;  // master convention
             }
         }
@@ -743,7 +744,7 @@ function base64_decode(str) {
 
 function _utf8_decode(bytes) {
     // Convert a UTF-8 byte array back to a JavaScript string.
-    // Handles BMP codepoints which covers the text LivePilot exchanges.
+    // Handles BMP codepoints and 4-byte sequences (emoji/supplementary planes).
     var result = "";
     for (var i = 0; i < bytes.length;) {
         var b0 = bytes[i];
@@ -754,17 +755,30 @@ function _utf8_decode(bytes) {
             var b1 = bytes[i + 1];
             result += String.fromCharCode(((b0 & 0x1F) << 6) | (b1 & 0x3F));
             i += 2;
-        } else if (i + 2 < bytes.length) {
-            var b2 = bytes[i + 1];
-            var b3 = bytes[i + 2];
+        } else if ((b0 & 0xF0) === 0xE0 && i + 2 < bytes.length) {
+            // 3-byte sequence (U+0800..U+FFFF)
+            var b1_3 = bytes[i + 1];
+            var b2_3 = bytes[i + 2];
             result += String.fromCharCode(
                 ((b0 & 0x0F) << 12) |
-                ((b2 & 0x3F) << 6) |
-                (b3 & 0x3F)
+                ((b1_3 & 0x3F) << 6) |
+                (b2_3 & 0x3F)
             );
             i += 3;
+        } else if ((b0 & 0xF8) === 0xF0 && i + 3 < bytes.length) {
+            // 4-byte sequence (U+10000..U+10FFFF) — emoji and supplementary planes
+            var cp = ((b0 & 0x07) << 18) |
+                     ((bytes[i + 1] & 0x3F) << 12) |
+                     ((bytes[i + 2] & 0x3F) << 6) |
+                     (bytes[i + 3] & 0x3F);
+            // Encode as UTF-16 surrogate pair
+            cp -= 0x10000;
+            result += String.fromCharCode(0xD800 + (cp >> 10));
+            result += String.fromCharCode(0xDC00 + (cp & 0x3FF));
+            i += 4;
         } else {
-            break;
+            // Skip invalid byte
+            i += 1;
         }
     }
     return result;
@@ -1311,7 +1325,9 @@ function cmd_get_plugin_params(args) {
     var is_plugin = (class_name === "PluginDevice" || class_name === "AuPluginDevice");
     if (!is_plugin) {
         send_response({
-            "error": "Device is " + class_name + ", not a plugin (PluginDevice/AuPluginDevice)"
+            "error": "Device is " + class_name + ", not a plugin (PluginDevice/AuPluginDevice). " +
+                "This tool only works on AU/VST plugins. Use get_device_parameters for native Ableton devices. " +
+                "Check get_device_info().is_plugin to verify before calling."
         });
         return;
     }
@@ -1427,7 +1443,8 @@ function cmd_get_plugin_presets(args) {
     var is_plugin = (class_name === "PluginDevice" || class_name === "AuPluginDevice");
     if (!is_plugin) {
         send_response({
-            "error": "Device is " + class_name + ", not a plugin"
+            "error": "Device is " + class_name + ", not a plugin. " +
+                "This tool only works on AU/VST plugins. Check get_device_info().is_plugin first."
         });
         return;
     }

package/mcp_server/__init__.py

@@ -1,2 +1,2 @@
 """LivePilot MCP Server — bridges MCP protocol to Ableton Live."""
-__version__ = "1.9.14"
+__version__ = "1.9.16"

package/mcp_server/connection.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import asyncio
 import json
 import os
 import socket
@@ -40,6 +41,8 @@ def _friendly_error(code: str, message: str, command_type: str) -> str:
     """Format an error from the Remote Script into a user-friendly message."""
     hint = _ERROR_HINTS.get(code, "")
     parts = [f"[{code}] {message}"]
+    if command_type:
+        parts.append(f"(while running '{command_type}')")
     if hint:
         parts.append(hint)
     return " ".join(parts)
@@ -143,7 +146,9 @@ class AbletonConnection:
 
         Thread-safe: a lock serializes all TCP send/receive cycles to
         prevent socket corruption when multiple MCP tools fire concurrently.
-        Retries once on socket errors with a fresh connection.
+        Retries once on connection errors (command never reached Ableton).
+        Does NOT retry on timeouts — Ableton may have already processed the
+        command, and retrying would cause duplicate mutations.
         """
         with self._lock:
             # Ensure we have a connection
@@ -156,7 +161,15 @@ class AbletonConnection:
 
             try:
                 response = self._send_raw(command)
-            except (OSError, AbletonConnectionError):
+            except AbletonConnectionError as exc:
+                # Don't retry timeouts — Ableton may have processed the command
+                if "Timeout" in str(exc):
+                    raise
+                # Retry once with a fresh connection for non-timeout errors
+                self.disconnect()
+                self.connect()
+                response = self._send_raw(command)
+            except OSError:
                 # Retry once with a fresh connection
                 self.disconnect()
                 self.connect()
@@ -183,6 +196,15 @@ class AbletonConnection:
         self._command_log.append(log_entry)
         return response.get("result", {})
 
+    async def send_command_async(self, command_type: str, params: Optional[dict] = None) -> dict:
+        """Async wrapper around send_command that avoids blocking the event loop.
+
+        Runs the blocking TCP send/receive in a thread pool executor so the
+        asyncio event loop remains responsive to other concurrent MCP tools.
+        """
+        loop = asyncio.get_running_loop()
+        return await loop.run_in_executor(None, self.send_command, command_type, params)
+
     # ------------------------------------------------------------------
     # Command log
     # ------------------------------------------------------------------

package/mcp_server/curves.py

@@ -277,10 +277,18 @@ def _square(duration: float, density: int, low: float, high: float,
     return points
 
 
-def _steps(values: list[float], duration: float, **_) -> list:
-    """Quantized staircase from explicit value list."""
+def _steps(values: list[float], duration: float, start: float = 0.0,
+           end: float = 1.0, steps: int = 16, density: int = 16,
+           **_) -> list:
+    """Quantized staircase from explicit value list or auto-generated from start/end.
+
+    If values is empty, generates a staircase with `steps` evenly spaced
+    levels from `start` to `end`.
+    """
     if not values:
-        return []
+        # Auto-generate staircase from start/end with the given number of steps
+        n = max(steps, 2)
+        values = [start + (end - start) * i / (n - 1) for i in range(n)]
     step_dur = duration / len(values)
     return [
         {"time": i * step_dur, "value": v, "duration": step_dur}
@@ -306,7 +314,7 @@ def _perlin(duration: float, density: int, center: float = 0.5,
 
     def _hash_float(x: float, s: float) -> float:
         """Deterministic pseudo-random float from position + seed."""
-        h = hashlib.md5(f"{x:.6f}:{s:.6f}".encode()).hexdigest()
+        h = hashlib.md5(f"{x:.6f}:{s:.6f}".encode(), usedforsecurity=False).hexdigest()
         return (int(h[:8], 16) / 0xFFFFFFFF) * 2.0 - 1.0
 
     def _smoothstep(t: float) -> float:
@@ -355,7 +363,7 @@ def _brownian(duration: float, density: int, start: float = 0.5,
     import hashlib
 
     def _det_random(i: int, s: float) -> float:
-        h = hashlib.md5(f"{i}:{s:.6f}".encode()).hexdigest()
+        h = hashlib.md5(f"{i}:{s:.6f}".encode(), usedforsecurity=False).hexdigest()
         return (int(h[:8], 16) / 0xFFFFFFFF) * 2.0 - 1.0
 
     points = []
@@ -581,7 +589,7 @@ def _stochastic(duration: float, density: int, center: float = 0.5,
     import hashlib
 
     def _det_random(i: int, s: float) -> float:
-        h = hashlib.md5(f"{i}:{s:.6f}".encode()).hexdigest()
+        h = hashlib.md5(f"{i}:{s:.6f}".encode(), usedforsecurity=False).hexdigest()
         return (int(h[:8], 16) / 0xFFFFFFFF) * 2.0 - 1.0
 
     points = []

package/mcp_server/evaluation/__init__.py

@@ -0,0 +1 @@
+"""Evaluation Fabric — unified evaluation layer for all engines."""