ai-consultants 2.10.3 → 2.10.5

package/CLAUDE.md

@@ -6,7 +6,7 @@ AI Consultants is a multi-model AI deliberation system that queries up to 15 AI
 
 **Self-Exclusion**: The invoking agent is automatically excluded from the panel to prevent self-consultation. Claude Code won't query Claude, Codex CLI won't query Codex, etc.
 
-**Version**: 2.10.3
+**Version**: 2.10.5
 
 ## Distribution
 
@@ -457,9 +457,9 @@ All consultants now use premium models by default:
 | Mistral | mistral-large-3 |
 | Cursor | composer-1.5 |
 | DeepSeek | deepseek-reasoner |
-| GLM | glm-5 |
+| GLM | glm-5.1 |
 | Grok | grok-4.20-0309-reasoning |
-| Qwen3 | qwen3.5-plus |
+| Qwen3 | qwen3.6-plus |
 | Kimi | kimi-code/kimi-for-coding |
 | Aider | gpt-5.3-codex |
 | MiniMax | MiniMax-M2.7 |
@@ -728,6 +728,10 @@ For detailed information, see:
 - Use `.env.example` as template for environment configuration
 - Run `./scripts/doctor.sh` to verify configuration
 
+## Git Conventions
+
+Use [Conventional Commits](https://www.conventionalcommits.org/): `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`, `test:`, `perf:`, `ci:`. Scope is optional (e.g., `feat(routing): add fallback escalation`).
+
 ## Release Process
 
 Every version bump **must** include a release note in `docs/releases/v<VERSION>.md`. Use the template below.
@@ -798,6 +802,21 @@ curl -fsSL https://raw.githubusercontent.com/matteoscurati/ai-consultants/main/s
 
 ## Changelog
 
+### v2.10.5
+- Qwen3 premium model upgraded from `qwen3.5-plus` to `qwen3.6-plus` ($0.325/$1.95 per M tokens)
+- Qwen3 standard tier now uses open-weight `qwen3.6-35b-a3b` (MoE, 35B total / 3B active)
+- Refactored `get_economic_model()` to delegate to `get_model_for_tier()`, eliminating stale hardcoded mappings
+- Moved `qwen3.5-plus` to legacy section in cost catalog
+- Fixed `AI_CONSULTANTS_VERSION` in `config.sh` (was stuck at `2.10.0`)
+
+### v2.10.4
+- GLM premium/standard model upgraded from `glm-5` to `glm-5.1`
+- Fixed Kilo CLI hanging indefinitely in non-TTY mode (query via stdin instead of CLI argument)
+- Fixed Kilo CLI picking wrong provider when other consultants' API keys were in the environment
+- Collapsed 5-stage ANSI stripping pipeline into single sed invocation in `query_kilo.sh`
+- Fixed overly aggressive markdown fence filter in `query_kilo.sh` (only strips standalone ``` lines now)
+- Updated `.env.example` GLM signup URL to `open.z.ai`
+
 ### v2.10.3
 - Grok premium model upgraded to `grok-4.20-0309-reasoning` (replaces `grok-4-1-fast-reasoning`)
 - GLM API endpoint migrated from `open.bigmodel.cn` to `api.z.ai/api/coding/paas/v4`

package/README.md

@@ -1,8 +1,8 @@
-# AI Consultants v2.10.3
+# AI Consultants v2.10.5
 
 > Query multiple AI models simultaneously for expert opinions on coding questions. Get diverse perspectives, automatic synthesis, confidence-weighted recommendations, and multi-agent debate.
 
-[![Version](https://img.shields.io/badge/version-2.10.3-blue.svg)](https://github.com/matteoscurati/ai-consultants)
+[![Version](https://img.shields.io/badge/version-2.10.5-blue.svg)](https://github.com/matteoscurati/ai-consultants)
 [![License](https://img.shields.io/badge/license-MIT-green.svg)](LICENSE)
 [![Claude Code Skill](https://img.shields.io/badge/Claude%20Code-Skill-orange.svg)](https://docs.anthropic.com/en/docs/claude-code/skills)
 [![GitHub stars](https://img.shields.io/github/stars/matteoscurati/ai-consultants?style=social)](https://github.com/matteoscurati/ai-consultants)
@@ -419,8 +419,8 @@ INVOKING_AGENT=codex ./scripts/consult_all.sh "Question"    # Codex excluded
 
 | Consultant | Default Model | Persona | Focus |
 |------------|---------------|---------|-------|
-| **Qwen3** | qwen3.5-plus | The Analyst | Data-driven analysis |
-| **GLM** | glm-5 | The Methodologist | Structured approaches |
+| **Qwen3** | qwen3.6-plus | The Analyst | Data-driven analysis |
+| **GLM** | glm-5.1 | The Methodologist | Structured approaches |
 | **Grok** | grok-4.20-0309-reasoning | The Provocateur | Challenge conventions |
 | **DeepSeek** | deepseek-reasoner | The Code Specialist | Algorithms, code generation |
 | **MiniMax** | MiniMax-M2.7 | The Pragmatic Optimizer | Performance, efficiency, pragmatism |
@@ -477,9 +477,9 @@ Choose the right balance of quality, speed, and cost with model quality tiers.
 | Mistral | mistral-large-3 | mistral-medium-latest | devstral-small-2 |
 | Cursor | composer-1.5 | composer-1.5 | gemini-2.0-flash |
 | DeepSeek | deepseek-reasoner | deepseek-v3.2 | deepseek-chat |
-| GLM | glm-5 | glm-5 | glm-4-flash |
+| GLM | glm-5.1 | glm-5.1 | glm-4-flash |
 | Grok | grok-4.20-0309-reasoning | grok-3 | grok-3-mini |
-| Qwen3 | qwen3.5-plus | qwen3.5-plus | qwen3-32b |
+| Qwen3 | qwen3.6-plus | qwen3.6-35b-a3b | qwen3-32b |
 | Aider | gpt-5.3-codex | gpt-5.3 | gpt-4o-mini |
 | MiniMax | MiniMax-M2.7 | MiniMax-M2.7 | MiniMax-M2.5 |
 | Ollama | qwen2.5-coder:32b | llama3.3 | llama3.2 |

package/SKILL.md

@@ -5,10 +5,10 @@ license: MIT
 compatibility: Requires bash, jq, and at least 2 AI CLI tools (gemini, codex, vibe, etc.). macOS and Linux.
 metadata:
   author: matteoscurati
-  version: 2.10.3
+  version: 2.10.5
 ---
 
-# AI Consultants v2.10.3 - AI Expert Panel
+# AI Consultants v2.10.5 - AI Expert Panel
 
 Simultaneously consult multiple AIs as "consultants" for coding questions. Each consultant has a **configurable persona** that influences their response style.
 

package/docs/COST_RATES.md

@@ -1,4 +1,4 @@
-# Cost Rates - AI Consultants v2.10.3
+# Cost Rates - AI Consultants v2.10.5
 
 This page documents the per-token rates used by the cost tracking system.
 
@@ -34,12 +34,13 @@ Since v2.5, models are organized into three tiers. Use `apply_model_tier()` or `
 | deepseek-reasoner | Premium | $0.002 | DeepSeek |
 | deepseek-v3.2 | Standard | $0.0014 | DeepSeek |
 | deepseek-chat | Economy | $0.001 | DeepSeek |
-| glm-5 | Premium/Standard | $0.002 | GLM |
+| glm-5.1 | Premium/Standard | $0.002 | GLM |
 | glm-4-flash | Economy | $0.0005 | GLM |
 | grok-4.20-0309-reasoning | Premium | $0.005 | Grok |
 | grok-3 | Standard | $0.003 | Grok |
 | grok-3-mini | Economy | $0.001 | Grok |
-| qwen3.5-plus | Premium/Standard | $0.002 | Qwen3 |
+| qwen3.6-plus | Premium | $0.000325 | Qwen3 |
+| qwen3.6-35b-a3b | Standard | $0.000163 | Qwen3 |
 | qwen3-32b | Economy | $0.0005 | Qwen3 |
 | MiniMax-M2.7 | Premium/Standard | $0.002 | MiniMax |
 | MiniMax-M2.5 | Economy | $0.0015 | MiniMax |
@@ -67,12 +68,13 @@ Since v2.5, models are organized into three tiers. Use `apply_model_tier()` or `
 | deepseek-reasoner | Premium | $0.006 | DeepSeek |
 | deepseek-v3.2 | Standard | $0.004 | DeepSeek |
 | deepseek-chat | Economy | $0.003 | DeepSeek |
-| glm-5 | Premium/Standard | $0.006 | GLM |
+| glm-5.1 | Premium/Standard | $0.006 | GLM |
 | glm-4-flash | Economy | $0.0015 | GLM |
 | grok-4.20-0309-reasoning | Premium | $0.015 | Grok |
 | grok-3 | Standard | $0.009 | Grok |
 | grok-3-mini | Economy | $0.003 | Grok |
-| qwen3.5-plus | Premium/Standard | $0.006 | Qwen3 |
+| qwen3.6-plus | Premium | $0.00195 | Qwen3 |
+| qwen3.6-35b-a3b | Standard | $0.0009 | Qwen3 |
 | qwen3-32b | Economy | $0.0015 | Qwen3 |
 | MiniMax-M2.7 | Premium/Standard | $0.006 | MiniMax |
 | MiniMax-M2.5 | Economy | $0.004 | MiniMax |

package/docs/cost_rates.json

@@ -1,5 +1,5 @@
 {
-  "version": "2.10.3",
+  "version": "2.10.5",
   "updated": "2026-03-04",
   "models": {
     "_comment_premium": "Premium tier models - March 2026",
@@ -8,9 +8,10 @@
     "gpt-5.3-codex": {"input": 3.00, "output": 12.00},
     "mistral-large-3": {"input": 2.00, "output": 6.00},
     "deepseek-reasoner": {"input": 0.55, "output": 2.19},
-    "glm-5": {"input": 0.50, "output": 2.00},
+    "glm-5.1": {"input": 0.50, "output": 2.00},
     "grok-4.20-0309-reasoning": {"input": 5.00, "output": 15.00},
-    "qwen3.5-plus": {"input": 1.60, "output": 6.40},
+    "qwen3.6-plus": {"input": 0.325, "output": 1.95},
+    "qwen3.6-35b-a3b": {"input": 0.163, "output": 0.90},
     "composer-1.5": {"input": 0.005, "output": 0.015},
     "minimax-m2.7": {"input": 0.50, "output": 2.00},
     "kimi-code/kimi-for-coding": {"input": 0.50, "output": 2.00},
@@ -36,6 +37,7 @@
     "minimax-m2.5": {"input": 0.30, "output": 1.50},
 
     "_comment_legacy": "Legacy models for backwards compatibility",
+    "qwen3.5-plus": {"input": 1.60, "output": 6.40},
     "claude-opus-4-5-20251124": {"input": 5.00, "output": 25.00},
     "claude-sonnet-4-5-20251124": {"input": 3.00, "output": 15.00},
     "claude-3-5-haiku-20241022": {"input": 0.001, "output": 0.005},
@@ -73,9 +75,9 @@
     "codex": "gpt-5.3-codex",
     "mistral": "mistral-large-3",
     "deepseek": "deepseek-reasoner",
-    "glm": "glm-5",
+    "glm": "glm-5.1",
     "grok": "grok-4.20-0309-reasoning",
-    "qwen3": "qwen3.5-plus",
+    "qwen3": "qwen3.6-plus",
     "aider": "gpt-5.3-codex",
     "kilo": "auto",
     "cursor": "composer-1.5",
@@ -92,9 +94,9 @@
       "codex": "gpt-5.3-codex",
       "mistral": "mistral-large-3",
       "deepseek": "deepseek-reasoner",
-      "glm": "glm-5",
+      "glm": "glm-5.1",
       "grok": "grok-4.20-0309-reasoning",
-      "qwen3": "qwen3.5-plus",
+      "qwen3": "qwen3.6-plus",
       "aider": "gpt-5.3-codex",
       "kilo": "auto",
       "cursor": "composer-1.5",
@@ -109,9 +111,9 @@
       "codex": "gpt-5.3",
       "mistral": "mistral-medium-latest",
       "deepseek": "deepseek-v3.2",
-      "glm": "glm-5",
+      "glm": "glm-5.1",
       "grok": "grok-3",
-      "qwen3": "qwen3.5-plus",
+      "qwen3": "qwen3.6-35b-a3b",
       "aider": "gpt-5.3",
       "kilo": "auto",
       "cursor": "composer-1.5",

package/docs/releases/v2.10.4.md

@@ -0,0 +1,63 @@
+# Release v2.10.4
+
+**Date:** 2026-04-10
+**Type:** Patch — GLM model upgrade, Kilo CLI fix
+**Previous:** v2.10.3
+
+## Highlights
+
+- GLM premium/standard model upgraded from `glm-5` to `glm-5.1`
+- Fixed Kilo CLI consultant that hung indefinitely when stdout was redirected
+- Fixed Kilo CLI auto-detecting other consultants' API keys and picking non-functional models
+- Improved output stripping pipeline efficiency in `query_kilo.sh`
+
+## What's New
+
+### GLM Model Upgrade
+
+GLM premium and standard tier updated from `glm-5` to `glm-5.1`. Economy tier (`glm-4-flash`) remains unchanged.
+
+| Tier | Before (v2.10.3) | After (v2.10.4) |
+|------|-------------------|------------------|
+| Premium | glm-5 | glm-5.1 |
+| Standard | glm-5 | glm-5.1 |
+| Economy | glm-4-flash | glm-4-flash |
+
+### Kilo CLI Fix
+
+Resolved a long-standing issue where the Kilo consultant would hang for the full timeout (180s) and return empty or ANSI-only output. Root causes:
+
+1. **Non-TTY hang**: Kilocode CLI hangs when the query is passed as a CLI argument and stdout is redirected. Fixed by passing the query via stdin instead.
+
+2. **Wrong provider auto-detection**: Kilocode auto-detects API providers from environment variables. When other consultants' API keys (e.g., `MINIMAX_API_KEY`) were exported, kilocode would pick non-functional models like `MiniMax-M2.7-highspeed`. Fixed by unsetting foreign API keys via `env -u` before invocation.
+
+3. **Output cleaning**: Collapsed 5-stage pipe (`sed | sed | tr | grep | grep`) into a single `sed` invocation. Fixed overly aggressive markdown fence filter that would strip opening code fences like `` ```typescript `` — now only strips standalone `` ``` `` lines.
+
+### Documentation Fixes
+
+- Updated `.env.example` header from v2.10.0 to current version
+- Updated GLM signup URL from `open.bigmodel.cn` to `open.z.ai`
+
+## Breaking Changes
+
+None.
+
+## Upgrade Guide
+
+```bash
+# If installed via git clone
+cd ~/.claude/skills/ai-consultants && git pull
+
+# If installed via curl | bash
+curl -fsSL https://raw.githubusercontent.com/matteoscurati/ai-consultants/main/scripts/install.sh | bash
+```
+
+No configuration changes required. Existing `GLM_MODEL` overrides continue to work.
+
+## Commits
+
+- `09dac4d` feat: upgrade GLM to glm-5.1, fix Kilo CLI hanging in non-TTY mode
+
+## Contributors
+
+- @matteoscurati

package/docs/releases/v2.10.5.md

@@ -0,0 +1,57 @@
+# Release v2.10.5
+
+**Date:** 2026-04-16
+**Type:** Patch — Qwen3 model upgrade to Qwen3.6
+**Previous:** v2.10.4
+
+## Highlights
+
+- Qwen3 premium default upgraded from `qwen3.5-plus` to `qwen3.6-plus` (80% cheaper)
+- Qwen3 standard tier now uses the open-weight `qwen3.6-35b-a3b` (MoE, 35B total / 3B active, Apache 2.0)
+- Eliminated stale hardcoded model mappings in `get_economic_model()`
+
+## What's New
+
+### Qwen3.6 Model Upgrade
+
+Alibaba released Qwen3.6-35B-A3B on April 16, 2026 — a sparse Mixture-of-Experts model with 35B total parameters but only 3B active during inference. The model scores 51.5 on Terminal-Bench 2.0 and 73.4 on SWE-bench Verified.
+
+| Tier | Before (v2.10.4) | After (v2.10.5) |
+|------|-------------------|------------------|
+| Premium | qwen3.5-plus ($1.60/$6.40 per M) | qwen3.6-plus ($0.325/$1.95 per M) |
+| Standard | qwen3.5-plus ($1.60/$6.40 per M) | qwen3.6-35b-a3b ($0.163/$0.90 per M) |
+| Economy | qwen3-32b | qwen3-32b (unchanged) |
+
+The premium tier (`qwen3.6-plus`) is Alibaba's hosted API with 1M context window. The standard tier (`qwen3.6-35b-a3b`) is the open-weight variant, also available on Ollama as `qwen3.6:35b-a3b` (24GB).
+
+### DRY Refactor: get_economic_model()
+
+`get_economic_model()` in `lib/costs.sh` had a hardcoded case statement with stale model names (e.g., `qwen-turbo`, `claude-3-haiku`, `gemini-2.5-flash`). Refactored to delegate to `get_model_for_tier()` in `config.sh`, matching the pattern already used by `get_premium_model()` in `lib/routing.sh`.
+
+### Version Fix
+
+Fixed `AI_CONSULTANTS_VERSION` in `config.sh` which was stuck at `2.10.0` since the npm packaging release.
+
+## Breaking Changes
+
+None.
+
+## Upgrade Guide
+
+```bash
+# If installed via git clone
+cd ~/.claude/skills/ai-consultants && git pull
+
+# If installed via curl | bash
+curl -fsSL https://raw.githubusercontent.com/matteoscurati/ai-consultants/main/scripts/install.sh | bash
+```
+
+No configuration changes required. Existing `QWEN3_MODEL` overrides continue to work.
+
+## Commits
+
+- `aaa1d69` feat(qwen3): upgrade to Qwen3.6-Plus and Qwen3.6-35B-A3B
+
+## Contributors
+
+- @matteoscurati

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-consultants",
-  "version": "2.10.3",
+  "version": "2.10.5",
   "description": "Multi-model AI deliberation system - query up to 15 AI consultants for coding problems",
   "bin": {
     "ai-consultants": "bin/ai-consultants"

package/references/configuration.md

@@ -79,8 +79,8 @@ CURSOR_MODEL=composer-1.5
 AIDER_MODEL=gpt-5.3-codex
 AMP_MODEL=amp
 KIMI_MODEL=kimi-code/kimi-for-coding
-QWEN3_MODEL=qwen3.5-plus
-GLM_MODEL=glm-5
+QWEN3_MODEL=qwen3.6-plus
+GLM_MODEL=glm-5.1
 GROK_MODEL=grok-4.20-0309-reasoning
 DEEPSEEK_MODEL=deepseek-reasoner
 MINIMAX_MODEL=MiniMax-M2.7

package/scripts/config.sh

@@ -105,7 +105,7 @@ AIDER_CMD="${AIDER_CMD:-aider}"
 # QWEN3 CONFIGURATION - The Analyst (CLI/API switchable v2.7)
 # =============================================================================
 
-QWEN3_MODEL="${QWEN3_MODEL:-qwen3.5-plus}"
+QWEN3_MODEL="${QWEN3_MODEL:-qwen3.6-plus}"
 QWEN3_TIMEOUT_SECONDS="${QWEN3_TIMEOUT:-180}"
 QWEN3_CMD="${QWEN3_CMD:-qwen}"
 QWEN3_API_URL="${QWEN3_API_URL:-https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation}"
@@ -117,7 +117,7 @@ QWEN3_FORMAT="${QWEN3_FORMAT:-qwen}"
 # GLM CONFIGURATION - The Code Specialist (API-based)
 # =============================================================================
 
-GLM_MODEL="${GLM_MODEL:-glm-5}"
+GLM_MODEL="${GLM_MODEL:-glm-5.1}"
 GLM_TIMEOUT_SECONDS="${GLM_TIMEOUT:-180}"
 GLM_API_URL="${GLM_API_URL:-https://api.z.ai/api/coding/paas/v4/chat/completions}"
 GLM_FORMAT="${GLM_FORMAT:-openai}"
@@ -531,9 +531,9 @@ get_model_for_tier() {
                 mistral)  echo "mistral-large-3" ;;
                 cursor)   echo "composer-1.5" ;;
                 deepseek) echo "deepseek-reasoner" ;;
-                glm)      echo "glm-5" ;;
+                glm)      echo "glm-5.1" ;;
                 grok)     echo "grok-4.20-0309-reasoning" ;;
-                qwen3)    echo "qwen3.5-plus" ;;
+                qwen3)    echo "qwen3.6-plus" ;;
                 aider)    echo "gpt-5.3-codex" ;;
                 amp)      echo "amp" ;;
                 ollama)   echo "qwen2.5-coder:32b" ;;
@@ -551,9 +551,9 @@ get_model_for_tier() {
                 mistral)  echo "mistral-medium-latest" ;;
                 cursor)   echo "composer-1.5" ;;  # Same as premium (single model)
                 deepseek) echo "deepseek-v3.2" ;;
-                glm)      echo "glm-5" ;;  # Same as premium (no mid-tier GLM)
+                glm)      echo "glm-5.1" ;;  # Same as premium (no mid-tier GLM)
                 grok)     echo "grok-3" ;;
-                qwen3)    echo "qwen3.5-plus" ;;  # Same as premium (single tier available)
+                qwen3)    echo "qwen3.6-35b-a3b" ;;  # Open-weight MoE (35B total, 3B active)
                 aider)    echo "gpt-5.3" ;;
                 amp)      echo "amp" ;;  # Same model (no tiers)
                 ollama)   echo "llama3.3" ;;
@@ -750,4 +750,4 @@ EOF
 # VERSION
 # =============================================================================
 
-AI_CONSULTANTS_VERSION="2.10.0"
+AI_CONSULTANTS_VERSION="2.10.5"

package/scripts/lib/common.sh

@@ -124,12 +124,12 @@ run_with_timeout() {
 # =============================================================================
 
 check_command() {
-    local cmd=$1
+    local cmd_name=$1
     local name=$2
     local install_hint=$3
 
-    if ! command -v "$cmd" &> /dev/null; then
-        log_error "$name not found (command: $cmd)"
+    if ! command -v "$cmd_name" &> /dev/null; then
+        log_error "$name not found (command: $cmd_name)"
         if [[ -n "$install_hint" ]]; then
             log_info "Install with: $install_hint"
         fi
@@ -612,7 +612,7 @@ map_set() {
     local keys_var="_MAP_${map_name}__KEYS__"
     local current_keys
     eval "current_keys=\"\${$keys_var:-}\""
-    if [[ ! " $current_keys " =~ " $safe_key " ]]; then
+    if [[ " $current_keys " != *" $safe_key "* ]]; then
         if [[ -z "$current_keys" ]]; then
             eval "$keys_var=\"\$safe_key\""
         else

package/scripts/lib/costs.sh

@@ -476,26 +476,15 @@ get_model_tier() {
 }
 
 # Get economic model for a consultant
+# Delegates to get_model_for_tier() in config.sh (single source of truth)
 # Usage: get_economic_model <consultant>
 get_economic_model() {
     local consultant="$1"
-    consultant=$(echo "$consultant" | tr '[:upper:]' '[:lower:]')
-
-    case "$consultant" in
-        gemini)   echo "gemini-2.5-flash" ;;
-        codex)    echo "gpt-4o-mini" ;;
-        mistral)  echo "mistral-small" ;;
-        kilo)     echo "kilo" ;;           # No economy variant
-        cursor)   echo "cursor" ;;          # No economy variant
-        claude)   echo "claude-3-haiku" ;;
-        qwen3)    echo "qwen-turbo" ;;
-        glm)      echo "glm-3-turbo" ;;
-        grok)     echo "grok-3-mini" ;;
-        deepseek) echo "deepseek-chat" ;;
-
-        minimax)  echo "MiniMax-M2.5" ;;
-        *)        echo "" ;;
-    esac
+    if type get_model_for_tier &>/dev/null; then
+        get_model_for_tier "$consultant" "economy"
+    else
+        echo ""
+    fi
 }
 
 # Calculate query complexity score (1-10)

package/scripts/lib/routing.sh

@@ -257,13 +257,15 @@ get_selected_consultants_json() {
     local category="$1"
     local min_affinity="${2:-7}"
 
-    local consultants=$(select_consultants "$category" "$min_affinity")
+    local consultants
+    consultants=$(select_consultants "$category" "$min_affinity")
     local json="["
     local first=true
 
     while IFS= read -r c; do
         if [[ -n "$c" ]]; then
-            local score=$(get_affinity "$category" "$c")
+            local score
+            score=$(get_affinity "$category" "$c")
             if [[ "$first" == "true" ]]; then
                 first=false
             else

package/scripts/query_kilo.sh

@@ -44,17 +44,22 @@ TEMP_OUTPUT=$(mktemp)
 TEMP_RAW=$(mktemp)
 trap 'rm -f "$TEMP_OUTPUT" "$TEMP_RAW"' EXIT
 
-# Build command args: run subcommand + --auto + --dir /tmp (prevent SKILL.md loading)
-KILO_ARGS=("$KILO_CMD" "run" "--auto" "--dir" "/tmp")
+log_info "Consulting Kilo Code (timeout: ${KILO_TIMEOUT_SECONDS}s)..."
+
+# Kilocode hangs when the query is passed as a CLI argument with stdout redirected.
+# Passing the query via stdin avoids this issue.
+# Also: unset API keys from other consultants — kilocode auto-detects providers
+# from env vars (e.g. MINIMAX_API_KEY) and may pick a non-functional model.
+KILO_STDIN_ARGS=("env"
+    "-u" "MINIMAX_API_KEY" "-u" "GLM_API_KEY" "-u" "GROK_API_KEY"
+    "-u" "DEEPSEEK_API_KEY" "-u" "OPENAI_API_KEY" "-u" "ANTHROPIC_API_KEY"
+    "-u" "GEMINI_API_KEY" "-u" "MISTRAL_API_KEY" "-u" "QWEN3_API_KEY"
+    "$KILO_CMD" "run" "--auto" "--dir" "/tmp")
 if [[ -n "${KILO_MODEL:-}" && "${KILO_MODEL}" != "auto" ]]; then
-    KILO_ARGS+=("--model" "$KILO_MODEL")
+    KILO_STDIN_ARGS+=("--model" "$KILO_MODEL")
 fi
-KILO_ARGS+=("$FULL_QUERY")
-
-# Use run_with_timeout for cross-platform support (Linux/macOS/POSIX)
-log_info "Consulting Kilo Code (timeout: ${KILO_TIMEOUT_SECONDS}s)..."
 
-if run_with_timeout "$KILO_TIMEOUT_SECONDS" "${KILO_ARGS[@]}" > "$TEMP_RAW" 2>&1; then
+if echo "$FULL_QUERY" | run_with_timeout "$KILO_TIMEOUT_SECONDS" "${KILO_STDIN_ARGS[@]}" > "$TEMP_RAW" 2>&1; then
     exit_code=0
 else
     exit_code=$?
@@ -62,8 +67,9 @@ fi
 
 # --- Strip ANSI escape codes from Kilo's text output ---
 if [[ -f "$TEMP_RAW" && -s "$TEMP_RAW" ]]; then
-    # Kilo outputs plain text with ANSI codes; strip them for clean output
-    CONTENT=$(LC_ALL=C sed 's/\x1b\[[0-9;]*[a-zA-Z]//g' "$TEMP_RAW" | sed '/^$/d') || true
+    # Strip ANSI codes, carriage returns, empty lines, kilo's model header, and markdown fences
+    CONTENT=$(LC_ALL=C sed -e 's/\x1b\[[0-9;]*[a-zA-Z]//g' -e 's/\r//g' -e '/^$/d' \
+        -e '/^> code · /d' -e '/^```$/d' "$TEMP_RAW") || true
 
     if [[ -n "${CONTENT:-}" ]]; then
         exit_code=0

package/scripts/query_qwen3.sh

@@ -4,7 +4,7 @@
 # Usage: ./query_qwen3.sh "question" [context_file] [output_file]
 #
 # Environment variables:
-#   QWEN3_MODEL     - Model to use (default: qwen3.5-plus)
+#   QWEN3_MODEL     - Model to use (default: qwen3.6-plus)
 #   QWEN3_TIMEOUT   - Timeout in seconds (default: 180)
 #   QWEN3_USE_API   - Use API mode instead of CLI (default: false)
 #   QWEN3_API_KEY   - API key for DashScope API (required for API mode)
@@ -82,7 +82,7 @@ END_TIME=$(get_timestamp_ms)
 LATENCY_MS=$((END_TIME - START_TIME))
 
 # --- Configuration for response building ---
-MODEL_USED="${QWEN3_MODEL:-qwen3.5-plus}"
+MODEL_USED="${QWEN3_MODEL:-qwen3.6-plus}"
 PERSONA_NAME=$(get_persona_name "$CONSULTANT_NAME")
 
 # --- Post-processing: use shared helper ---