refactor: full agent system revision — migrate to GLM-5.1, fix delegation chains, audit consistency

- Migrate 8 agents from openrouter/qwen3.6-plus:free to ollama-cloud/glm-5.1
- Assign thinking/variant/instant depth by role complexity
- Fix broken delegation chains: system-analyst, all developer agents, devops-engineer now can reach orchestrator
- Add task permissions to browser-automation, visual-tester, capability-analyst, markdown-validator
- Add visual-tester permission to flutter-developer and frontend-developer
- Fix capability-index.yaml routing map indentation (go_* keys misplaced)
- Add delegates_to and variant fields to capability-index.yaml
- Update KILO_SPEC.md agent table with Variant column
- Update AGENTS.md with Model/Variant/CanCall columns
- Update kilo.jsonc ask agent model
- Fix YAML indentation in capability-analyst.md and markdown-validator.md
- Update agent-architect.md template models (remove gpt-oss, qwen3.6-plus)
- Add Skills Reference tables to 7 previously unlinked agents
- Full audit: 10/10 consistency checks passed

.kilo/agents/agent-architect.md

@@ -1,7 +1,8 @@
 ---
 name: Agent Architect
 mode: subagent
-model: openrouter/qwen/qwen3.6-plus:free
+model: ollama-cloud/glm-5.1
+variant: thinking
 description: Creates, modifies, and reviews new agents, workflows, and skills based on capability gap analysis
 color: "#8B5CF6"
 permission:
@@ -61,10 +62,10 @@ Grant only required permissions:
 
 ### Appropriate Models
 Choose cost-effective models:
-- Complex reasoning: ollama-cloud/gpt-oss:120b
+- Complex reasoning: ollama-cloud/glm-5.1 (thinking)
 - Code generation: ollama-cloud/qwen3-coder:480b
-- Analysis: ollama-cloud/gpt-oss:120b
-- Simple tasks: qwen/qwen3.6-plus:free
+- Analysis: ollama-cloud/nemotron-3-super
+- Simple/validation: ollama-cloud/nematron-3-nano:30b
 
 ## Creation Process
 
@@ -287,7 +288,7 @@ cat > .kilo/agents/api-architect.md << 'EOF'
 ---
 description: Design and validate API schemas
 mode: subagent
-model: ollama-cloud/gpt-oss:120b
+model: ollama-cloud/glm-5.1
 color: "#F59E0B"
 permission:
   read: allow

.kilo/agents/backend-developer.md

@@ -13,6 +13,7 @@ permission:
   task:
     "*": deny
     "code-skeptic": allow
+    "orchestrator": allow
 ---
 
 # Kilo Code: Backend Developer

.kilo/agents/browser-automation.md

@@ -11,6 +11,9 @@ permission:
   glob: allow
   grep: allow
   webfetch: allow
+  task:
+    "*": deny
+    "orchestrator": allow
 ---
 
 # Kilo Code: Browser Automation Agent
@@ -19,6 +22,13 @@ permission:
 
 You are **Browser Automation Agent** — an expert in web testing and browser control via Playwright MCP. You can navigate pages, fill forms, click elements, take screenshots, and validate UI using the Model Context Protocol.
 
+## Skills Reference
+
+| Skill | Purpose |
+|-------|---------|
+| `playwright` | Playwright MCP setup and patterns |
+| `web-testing` | Web testing strategies |
+
 ## When to Use
 
 Invoke this agent when:

.kilo/agents/capability-analyst.md

@@ -1,8 +1,16 @@
 ---
 description: Analyzes task requirements against available agents, workflows, and skills. Identifies gaps and recommends new components.
 mode: subagent
-model: openrouter/qwen/qwen3.6-plus:free
+model: ollama-cloud/glm-5.1
 color: "#6366F1"
+permission:
+  read: allow
+  glob: allow
+  grep: allow
+  task:
+    "*": deny
+    "agent-architect": allow
+    "orchestrator": allow
 ---
 
 # Capability Analyst Agent

.kilo/agents/code-skeptic.md

@@ -132,6 +132,12 @@ Tests:
 - DO NOT focus only on style
 - DO NOT skip security review
 
+## Skills Reference
+
+| Skill | Purpose |
+|-------|---------|
+| `quality-controller` | Quality gate patterns and checklists |
+
 ## Handoff Protocol
 
 After review:

.kilo/agents/devops-engineer.md

@@ -14,6 +14,7 @@ permission:
     "*": deny
     "code-skeptic": allow
     "security-auditor": allow
+    "orchestrator": allow
 ---
 
 # Kilo Code: DevOps Engineer

.kilo/agents/evaluator.md

@@ -1,7 +1,8 @@
 ---
 description: Scores agent effectiveness after task completion for continuous improvement
 mode: subagent
-model: openrouter/qwen/qwen3.6-plus:free
+model: ollama-cloud/glm-5.1
+variant: thinking
 color: "#047857"
 permission:
   read: allow

.kilo/agents/flutter-developer.md

@@ -13,6 +13,8 @@ permission:
   task:
     "*": deny
     "code-skeptic": allow
+    "visual-tester": allow
+    "orchestrator": allow
 ---
 
 # Kilo Code: Flutter Developer

.kilo/agents/frontend-developer.md

@@ -13,6 +13,8 @@ permission:
   task:
     "*": deny
     "code-skeptic": allow
+    "visual-tester": allow
+    "orchestrator": allow
 ---
 
 # Kilo Code: Frontend Developer

.kilo/agents/go-developer.md

@@ -13,6 +13,7 @@ permission:
   task:
     "*": deny
     "code-skeptic": allow
+    "orchestrator": allow
 ---
 
 # Kilo Code: Go Developer

.kilo/agents/lead-developer.md

@@ -2,6 +2,7 @@
 description: Primary code writer for backend and core logic. Writes implementation to pass tests
 mode: subagent
 model: ollama-cloud/qwen3-coder:480b
+variant: thinking
 color: "#DC2626"
 permission:
   read: allow

.kilo/agents/markdown-validator.md

@@ -3,6 +3,14 @@ description: Validates and corrects Markdown descriptions for Gitea issues
 mode: subagent
 model: ollama-cloud/nemotron-3-nano:30b
 color: "#F97316"
+permission:
+  read: allow
+  edit: allow
+  glob: allow
+  grep: allow
+  task:
+    "*": deny
+    "orchestrator": allow
 ---
 
 # Markdown Validator Agent

.kilo/agents/memory-manager.md

@@ -18,6 +18,12 @@ permission:
 
 You are **Memory Manager** — responsible for managing all memory systems. Based on Lilian Weng's agent architecture research.
 
+## Skills Reference
+
+| Skill | Purpose |
+|-------|---------|
+| `memory-systems` | Memory architecture patterns |
+
 ## Memory Types
 
 ### 1. Short-Term Memory (Context Window)

.kilo/agents/orchestrator.md

@@ -1,7 +1,8 @@
 ---
 description: Main dispatcher. Routes tasks between agents based on Issue status and manages the workflow state machine. IF:90 for optimal routing accuracy.
 mode: all
-model: ollama-cloud/glm-5
+model: ollama-cloud/glm-5.1
+variant: thinking
 color: "#7C3AED"
 permission:
   read: allow

.kilo/agents/pipeline-judge.md

@@ -1,7 +1,7 @@
 ---
 description: Automated pipeline judge. Evaluates workflow execution by running tests, measuring token cost and wall-clock time. Produces objective fitness scores. Never writes code - only measures and scores.
 mode: subagent
-model: openrouter/qwen/qwen3.6-plus:free
+model: ollama-cloud/glm-5.1
 color: "#DC2626"
 permission:
   read: allow
@@ -35,6 +35,12 @@ You produce a **fitness score** that drives evolutionary optimization.
 - After a model swap recommendation is applied
 - On `/evaluate` command
 
+## Skills Reference
+
+| Skill | Purpose |
+|-------|---------|
+| `evolution-sync` | Fitness history synchronization and dashboard data |
+
 ## Fitness Score Formula
 
 ```

.kilo/agents/planner.md

@@ -53,3 +53,10 @@ Iterative execution with reflection between steps.
 ### Rollback Plan
 If {failure}: {rollback_action}
 ```
+
+## Skills Reference
+
+| Skill | Purpose |
+|-------|---------|
+| `planning-patterns` | CoT/ToT/Plan-Execute-Reflect strategies |
+| `task-analysis` | Task decomposition and dependency analysis |

.kilo/agents/product-owner.md

@@ -1,7 +1,7 @@
 ---
 description: Manages issue checklists, status labels, tracks progress and coordinates with human users
 mode: subagent
-model: openrouter/qwen/qwen3.6-plus:free
+model: ollama-cloud/glm-5.1
 color: "#EA580C"
 permission:
   read: allow
@@ -90,6 +90,13 @@ After update:
 2. Update checklist checkboxes
 3. Update status labels
 4. Notify relevant agents
+
+## Skills Reference
+
+| Skill | Purpose |
+|-------|---------|
+| `gitea` | Gitea API integration |
+| `scoped-labels` | Label management and status tracking |
 ## Gitea Commenting (MANDATORY)
 
 **You MUST post a comment to the Gitea issue after completing your work.**

.kilo/agents/prompt-optimizer.md

@@ -1,7 +1,8 @@
 ---
 description: Improves agent system prompts based on performance failures. Meta-learner for prompt optimization
 mode: subagent
-model: openrouter/qwen/qwen3.6-plus:free
+model: ollama-cloud/glm-5.1
+variant: instant
 color: "#BE185D"
 permission:
   read: allow

.kilo/agents/reflector.md

@@ -13,6 +13,12 @@ permission:
 
 # Kilo Code: Reflector
 
+## Skills Reference
+
+| Skill | Purpose |
+|-------|---------|
+| `research-cycle` | Self-improvement and research iteration patterns |
+
 ## Role Definition
 
 You are **Reflector** — the self-improvement specialist using Reflexion pattern (Shinn & Labash 2023).

.kilo/agents/release-manager.md

@@ -1,7 +1,7 @@
 ---
 description: Manages git operations, semantic versioning, branching, and deployments. Ensures clean history
 mode: subagent
-model: openrouter/qwen/qwen3.6-plus:free
+model: ollama-cloud/glm-5.1
 color: "#581C87"
 permission:
   read: allow

.kilo/agents/requirement-refiner.md

@@ -1,7 +1,8 @@
 ---
 description: Converts vague ideas and bug reports into strict User Stories with acceptance criteria checklists
 mode: all
-model: ollama-cloud/glm-5
+model: ollama-cloud/glm-5.1
+variant: thinking
 color: "#4F46E5"
 permission:
   read: allow

.kilo/agents/sdet-engineer.md

@@ -2,6 +2,7 @@
 description: Writes tests following TDD methodology. Tests MUST fail initially (Red phase)
 mode: all
 model: ollama-cloud/qwen3-coder:480b
+variant: thinking
 color: "#8B5CF6"
 permission:
   read: allow

.kilo/agents/system-analyst.md

@@ -1,7 +1,8 @@
 ---
 description: Designs technical specifications, data schemas, and API contracts before implementation
 mode: subagent
-model: ollama-cloud/glm-5
+model: ollama-cloud/glm-5.1
+variant: thinking
 color: "#0891B2"
 permission:
   read: allow
@@ -12,6 +13,8 @@ permission:
   grep: allow
   task:
     "*": deny
+    "sdet-engineer": allow
+    "orchestrator": allow
 ---
 
 # Kilo Code: System Analyst

.kilo/agents/the-fixer.md

@@ -96,6 +96,12 @@ After fix:
 2. Document the fix
 3. Use Task tool with subagent_type: "code-skeptic" for re-review
 4. If max iterations reached, use Task tool with subagent_type: "orchestrator" for escalation
+
+## Skills Reference
+
+| Skill | Purpose |
+|-------|---------|
+| `fix-workflow` | Iterative fix loop patterns |
 ## Gitea Commenting (MANDATORY)
 
 **You MUST post a comment to the Gitea issue after completing your work.**

.kilo/agents/visual-tester.md

@@ -10,6 +10,10 @@ permission:
   bash: allow
   glob: allow
   grep: allow
+  task:
+    "*": deny
+    "the-fixer": allow
+    "orchestrator": allow
 ---
 
 # Kilo Code: Visual Tester Agent

.kilo/agents/workflow-architect.md

@@ -1,7 +1,8 @@
 ---
 description: Creates and maintains workflow definitions with complete architecture, Gitea integration, and quality gates
 mode: subagent
-model: openrouter/qwen/qwen3.6-plus:free
+model: ollama-cloud/glm-5.1
+variant: thinking
 color: "#EC4899"
 permission:
   read: allow
@@ -27,6 +28,12 @@ You are the **Workflow Architect** — responsible for creating workflow definit
 4. Ensure complete, tested, documented delivery
 5. Can be handed to client independently
 
+## Skills Reference
+
+| Skill | Purpose |
+|-------|---------|
+| `gitea-workflow` | Gitea issue/label integration patterns |
+
 ## When to Use
 
 Invoke when: