opencode-enhancer-plugin 1.0.0 → 1.4.0

package/dist/index.js

@@ -1,29 +1,268 @@
+import { TodoParser } from "./utils/todoParser.js";
+import { todoStore } from "./utils/sessionStore.js";
+// === TODO ENFORCER CONSTANTS ===
+const TODO_ENFORCER_AGENTS = ['ultraplan', 'strategos'];
+const TODO_MARKER_START = '<!-- TODOS-START -->';
+const TODO_MARKER_END = '<!-- TODOS-END -->';
+// === STRATEGOS MODE CONSTANTS ===
+const STRATEGOS_KEYWORDS = [
+    'strategos',
+    'strategy',
+    'strat',
+    'interview mode',
+    'strategic plan',
+    'complex plan',
+    'architecture decision',
+    'design review',
+    'comprehensive plan',
+    '🎯',
+];
+// Validation Constants
+const VALID_MODEL_PATTERNS = [
+    /^opencode\//,
+    /^anthropic\//,
+    /^openai\//,
+    /^google\//,
+    /^mistral\//,
+    /^cohere\//,
+    /^ollama\//
+];
+const DEFAULT_MODEL = "opencode/kimi-k2.5-free";
+// Model Validation Function
+function validateModel(model, agentName, client) {
+    if (!model || typeof model !== 'string') {
+        client.app.log({
+            service: "enhancer",
+            level: "warn",
+            message: `Invalid model for ${agentName}, using default`
+        }).catch(() => { });
+        return DEFAULT_MODEL;
+    }
+    const isValid = VALID_MODEL_PATTERNS.some(pattern => pattern.test(model));
+    if (!isValid) {
+        client.app.log({
+            service: "enhancer",
+            level: "warn",
+            message: `Unknown model provider for ${agentName}: ${model}, using default`
+        }).catch(() => { });
+        return DEFAULT_MODEL;
+    }
+    return model;
+}
+// Model Resolver with Hierarchical Configuration
+function resolveModel(agentName, agentType, existingConfig, client) {
+    let model;
+    // 1. Check existing OpenCode Config (highest priority)
+    if (existingConfig?.model) {
+        return validateModel(existingConfig.model, agentName, client);
+    }
+    // 2. Specific environment variable (e.g., ENHANCER_MODEL_REVIEW_PLAN)
+    const specificVar = `ENHANCER_MODEL_${agentName.toUpperCase().replace(/-/g, '_')}`;
+    model = process.env[specificVar];
+    if (model) {
+        client.app.log({
+            service: "enhancer",
+            level: "info",
+            message: `Using specific env var ${specificVar} for ${agentName}`
+        }).catch(() => { });
+        return validateModel(model, agentName, client);
+    }
+    // 3. Grouped environment variable (ENHANCER_MODEL_SUBAGENT or ENHANCER_MODEL_PRIMARY)
+    const groupVar = `ENHANCER_MODEL_${agentType.toUpperCase()}`;
+    model = process.env[groupVar];
+    if (model) {
+        client.app.log({
+            service: "enhancer",
+            level: "info",
+            message: `Using group env var ${groupVar} for ${agentName}`
+        }).catch(() => { });
+        return validateModel(model, agentName, client);
+    }
+    // 4. Default
+    return DEFAULT_MODEL;
+}
 export const EnhancerPlugin = async ({ client }) => {
     client.app.log({
         service: "enhancer",
         level: "info",
-        message: "Enhancer plugin initialized"
+        message: "🔥 Enhancer plugin v1.4.0 initialized with Todo Enforcer & Strategos Mode"
     }).catch(() => { });
     return {
         config: async (input) => {
             if (!input.agent) {
                 input.agent = {};
             }
-            input.agent["explore-context"] = {
-                mode: "subagent",
-                hidden: true,
-                model: "opencode/kimi-k2.5-free",
-                prompt: `You are the UNIVERSAL CONTEXT DETECTIVE. Map the codebase and extract technical context.`,
+            // Helper function to configure agent with merging logic
+            const configureAgent = (name, type, defaultConfig) => {
+                const existing = input.agent[name] || {};
+                const resolvedModel = resolveModel(name, type, existing, client);
+                // Merge: Default → User Config → Resolved Model → Fixed Props
+                input.agent[name] = {
+                    ...defaultConfig,
+                    ...existing,
+                    model: resolvedModel,
+                    mode: type === 'subagent' ? 'subagent' : 'primary',
+                    ...(type === 'subagent' ? { hidden: true } : {})
+                };
+            };
+            // Configure Subagents
+            // explore-context
+            configureAgent('explore-context', 'subagent', {
+                prompt: `Detect project identity and tech stack. Output: [PROJECT_IDENTITY] with Type, Entry Point, Main Directories.`,
                 tools: {
                     list: true,
                     read: true,
                     grep: true,
                     bash: true
                 }
-            };
-            input.agent.enhancer = {
-                mode: "primary",
-                model: "opencode/kimi-k2.5-free",
+            });
+            // explore-code
+            configureAgent('explore-code', 'subagent', {
+                prompt: `Analyze source code files. Find relevant code for user request. Output: [CODE_ANALYSIS] with Relevant Files, Key Functions, Patterns.`,
+                tools: {
+                    list: true,
+                    read: true,
+                    grep: true,
+                    bash: true
+                }
+            });
+            // explore-deps
+            configureAgent('explore-deps', 'subagent', {
+                prompt: `Analyze dependencies and imports. Output: [DEPENDENCIES] with Core Libraries, Import Patterns, Context7 suggestions.`,
+                tools: {
+                    read: true,
+                    grep: true,
+                    bash: true
+                }
+            });
+            // explore-tests
+            configureAgent('explore-tests', 'subagent', {
+                prompt: `Analyze test structure. Output: [TEST_FRAMEWORK] with Framework, Config, Test Patterns, Existing Tests.`,
+                tools: {
+                    list: true,
+                    read: true,
+                    grep: true,
+                    bash: true
+                }
+            });
+            // review-plan
+            configureAgent('review-plan', 'subagent', {
+                prompt: `You are a PLAN REVIEWER. Your job is to critically analyze implementation plans for completeness, correctness, and feasibility.
+
+TASK: Review the provided plan and original context. Output a structured review report.
+
+REVIEW CRITERIA:
+1. Completeness: Does the plan cover all requirements? Are there missing steps?
+2. Correctness: Are the technical approaches sound? Are file paths accurate?
+3. Feasibility: Can this plan be executed successfully? Are there risks?
+4. Edge Cases: Are error scenarios and edge cases handled?
+5. Dependencies: Are all required dependencies identified and addressed?
+6. Testing: Is there a clear validation/testing strategy?
+
+OUTPUT FORMAT (Markdown):
+# Plan Review Report
+
+## Summary
+- Status: [APPROVED / NEEDS_REVISION]
+- Total Issues: [number]
+- Critical Issues: [number]
+- Warnings: [number]
+- Suggestions: [number]
+
+## Critical Issues (Must Fix)
+1. **[SEVERITY: Critical]** [Issue description] - [File/Area affected]
+   - Impact: [What could go wrong]
+   - Recommendation: [Specific fix]
+
+## Warnings (Should Fix)
+1. **[SEVERITY: Warning]** [Issue description] - [File/Area affected]
+   - Recommendation: [Specific improvement]
+
+## Suggestions (Nice to Have)
+1. **[SEVERITY: Suggestion]** [Issue description]
+   - Recommendation: [Enhancement idea]
+
+## Final Verdict
+[One paragraph summary of plan quality and readiness]`,
+                tools: {
+                    read: true,
+                    task: true
+                }
+            });
+            // strategos-interviewer (NEW)
+            configureAgent('strategos-interviewer', 'subagent', {
+                prompt: `You are STRATEGOS INTERVIEWER - a strategic requirements analyst.
+
+YOUR MISSION: Clarify unclear or complex requests through targeted interviewing.
+
+=== INTERVIEW METHODOLOGY ===
+
+1. **Request Analysis** (30 seconds)
+   - Identify the core problem
+   - Recognize ambiguities and assumptions
+   - Mark missing information
+
+2. **Clarification Questions** (2-5 questions)
+   Always ask about:
+   - ✓ **Scope**: What exactly is In/Out of scope?
+   - ✓ **Constraints**: Time, budget, technical limits?
+   - ✓ **Priorities**: What is most important? (Moscow: Must/Should/Could/Won't)
+   - ✓ **Success Criteria**: When is it "done"?
+   - ✓ **Edge Cases**: What could go wrong?
+   - ✓ **Dependencies**: What needs to happen first?
+   - ✓ **Stakeholders**: Who needs to be involved?
+   - ✓ **Integration**: Does it fit the existing architecture?
+
+3. **Follow-up questions** based on answers
+   - Go deeper on critical points
+   - Challenge assumptions
+   - Offer alternatives
+
+=== OUTPUT FORMAT ===
+
+# 🎯 Strategos Interview Report
+
+## Original Request
+[What is this about?]
+
+## Clarification Questions
+
+### Q1: [Question]
+**Context**: [Why is this important?]
+**Options**: [If multiple choice]
+
+### Q2: [Question]
+...
+
+### Q3: [Question]
+...
+
+## Anticipated Complexity
+- **Level**: [Low/Medium/High/Critical]
+- **Estimated Effort**: [X hours/days]
+- **Main Risks**: [List]
+
+## Recommended Approach
+1. [Suggestion 1]
+2. [Suggestion 2]
+
+---
+**Status**: ⏳ Waiting for user answers
+**Next Step**: Evaluate interview → Create plan
+
+=== RULES ===
+- NEVER ask more than 5 questions at once
+- Formulate questions to be easy to answer
+- Offer options for complex decisions
+- ALWAYS wait for user answers before proceeding`,
+                tools: {
+                    read: true,
+                    task: true,
+                },
+            });
+            // Configure Primary Agents
+            // enhancer
+            configureAgent('enhancer', 'primary', {
                 description: "Universal Technical Architect & Prompt Enhancer",
                 color: "#9C27B0",
                 steps: 15,
@@ -34,36 +273,557 @@ export const EnhancerPlugin = async ({ client }) => {
                     edit: false,
                     write: false
                 },
-                prompt: `You are the CHIEF ARCHITECT (Enhancer). Translate user wishes into precise engineering specifications.`
-            };
+                permission: {
+                    edit: "deny",
+                    bash: "deny"
+                },
+                prompt: `You are the CHIEF ARCHITECT (Enhancer). Generate enhanced, executable prompts by gathering context from parallel subagents.
+
+WORKFLOW:
+1. Classify intent: FIX/FEAT/REFACTOR/TEST/EXPLAIN
+2. Call MULTIPLE task tools IN PARALLEL based on intent:
+   - FIX/FEAT: explore-context + explore-code + explore-deps
+   - REFACTOR: explore-context + explore-code + explore-tests  
+   - TEST: explore-context + explore-code + explore-tests + explore-deps
+   - EXPLAIN: explore-context + explore-code
+3. Synthesize all results into unified understanding
+4. Generate enhanced markdown prompt with specific paths and patterns
+
+OUTPUT: Markdown code block with Task, Context, Instructions, Technical Constraints. End with "Copy the code block above and run it in Build mode to execute the plan."`
+            });
+            // ultraplan (UPDATED with Todo System)
+            configureAgent('ultraplan', 'primary', {
+                description: "Planning Agent with Discovery Subagents & Todo Enforcement 🔒",
+                color: "#FF5722",
+                steps: 20,
+                tools: {
+                    task: true,
+                    read: true,
+                    bash: false,
+                    edit: false,
+                    write: false
+                },
+                permission: {
+                    edit: "deny",
+                    bash: "deny"
+                },
+                prompt: `You are the ULTRAPLAN ARCHITECT - create implementation plans through discovery and analysis.
+
+=== STRICT WORKFLOW - FOLLOW EXACTLY ===
+
+**STEP 1: DISCOVERY PHASE**
+Call ALL explore-* subagents IN PARALLEL:
+- explore-context: Project structure and tech stack
+- explore-code: Relevant source code files  
+- explore-deps: Dependencies and imports
+- explore-tests: Test framework and existing tests
+
+WAIT for all subagents to return. Synthesize their findings.
+
+**STEP 2: TODO CREATION (CRITICAL!)**
+For every complex task, you MUST create a TODO list:
+
+## 📋 My TODOs
+- [ ] Step 1: [Concrete task]
+- [ ] Step 2: [Concrete task]
+- [ ] Step 3: [Concrete task]
+
+Format RULES:
+- ALWAYS use format: "- [ ] Description"
+- Mark completed with: "- [x] Description"
+- Be SPECIFIC (no vague tasks)
+- Create 3-10 todos depending on complexity
+
+**STEP 3: PLAN CREATION**
+Create a comprehensive implementation plan based on discovery results:
+- Specific file paths from actual analysis (NEVER invent paths)
+- Step-by-step implementation order
+- Error handling strategy
+- Validation/testing approach
+
+**STEP 4: OUTPUT**
+Output the plan as an executable markdown prompt with:
+- Clear task description
+- Context from discovery
+- Step-by-step instructions
+- Technical constraints
+- Your TODO list
+
+End with: "Copy the code block above and run it in **Build** mode to execute the plan."
+
+=== TODO ENFORCER SYSTEM 🔒 ===
+
+⚠️ **IMPORTANT**: You CANNOT stop until all Todos are marked with [x]!
+
+The system enforces:
+- Tracking of all your Todos
+- Blocking "Stop" with open Todos
+- Automatic continuation until everything is done
+
+Finish it! 🔥
+
+IMPORTANT RULES:
+- All file paths must come from actual code analysis, never invented
+- Never enable write/edit tools (analysis phase only)
+- Always use task tool for subagent calls
+
+=== WHEN TO USE STRATEGOS ===
+
+For complex tasks requiring interview mode, use Strategos (Tab → strategos):
+- New features (>1 file)
+- Architecture changes
+- Integrations
+- Performance/Security improvements`
+            });
+            // ask
+            configureAgent('ask', 'primary', {
+                description: "Context-Aware Question Answering Agent with Multi-Source Research",
+                color: "#2196F3",
+                steps: 15,
+                tools: {
+                    task: true,
+                    read: true,
+                    bash: false,
+                    edit: false,
+                    write: false
+                },
+                permission: {
+                    edit: "deny",
+                    bash: "deny"
+                },
+                prompt: `You are the ASK AGENT - a context-aware research assistant that answers questions by orchestrating parallel sub-agent investigations and synthesizing their findings into comprehensive answers.
+
+=== STRICT WORKFLOW - FOLLOW EXACTLY ===
+
+**STEP 1: QUESTION CLASSIFICATION**
+Analyze the user's question and classify into one of these types:
+- CODE_QUESTION: "How does X work?", "Explain this function", "What does this code do?", "Where is Y implemented?"
+- ARCHITECTURE_QUESTION: "How is this project structured?", "What patterns are used?", "What's the tech stack?"
+- DEBUG_QUESTION: "Why is this not working?", "Fix this error", "Why does this fail?"
+- GENERAL_INFO: "What is this project about?", "What does this repo do?"
+
+**STEP 2: PARALLEL SUB-AGENT ORCHESTRATION (CRITICAL - MUST BE PARALLEL)**
+Call MULTIPLE task tools IN PARALLEL based on question type:
+
+| Question Type | Sub-Agents to Call |
+|--------------|-------------------|
+| CODE_QUESTION | explore-context + explore-code + explore-deps |
+| ARCHITECTURE_QUESTION | explore-context + explore-code |
+| DEBUG_QUESTION | explore-context + explore-code + explore-deps + explore-tests |
+| GENERAL_INFO | explore-context |
+
+**STEP 3: WAIT AND SYNTHESIZE**
+- WAIT for ALL subagents to return (this is critical!)
+- Read returned context carefully
+- Synthesize information from all sources:
+  * Combine findings from different sub-agents
+  * Resolve any conflicts or gaps
+  * Identify the most relevant files and patterns
+  * Map relationships between components
+
+**STEP 4: ANSWER GENERATION**
+Generate a comprehensive answer with this exact structure:
+
+\`\`\`markdown
+## Answer
+
+[Clear, direct answer to the question - 2-4 sentences maximum]
+
+## Detailed Explanation
+
+[In-depth explanation with technical details]
+
+### Relevant Files
+- \`path/to/file.ts\` - [Why this file matters]
+- \`path/to/another.ts\` - [Why this file matters]
+
+### Key Code Sections
+\`\`\`typescript
+// From: path/to/file.ts (lines X-Y)
+[Most relevant code snippet]
+\`\`\`
+
+### Architecture Context
+- **Tech Stack**: [What technologies are used]
+- **Pattern**: [Design patterns identified]
+- **Dependencies**: [Key dependencies involved]
+
+## Related Areas to Explore
+- [Suggestion 1 with specific file path]
+- [Suggestion 2 with specific file path]
+\`\`\`
+
+**ERROR HANDLING RULES:**
+- If a sub-agent fails or times out: Continue with available context and note the limitation
+- If information is insufficient: Clearly state "Based on available context..." and explain what additional information would be needed
+- If conflicting information exists: Present both perspectives and indicate the conflict
+- Never invent file paths - only use paths returned by sub-agents
+
+**IMPORTANT:**
+- All file paths MUST come from actual sub-agent analysis
+- Use read tool to verify critical code snippets
+- Be concise but thorough - prioritize clarity over length
+- Always cite your sources (which sub-agent provided what information)`
+            });
+            // strategos (NEW - instead of Prometheus)
+            configureAgent('strategos', 'primary', {
+                description: "🎯 Strategic Planner with Interview Mode - Complex task planning through clarification",
+                color: "#FF6F00", // Deep Orange for Strategos
+                steps: 30,
+                tools: {
+                    task: true,
+                    read: true,
+                    bash: false,
+                    edit: false,
+                    write: false,
+                },
+                permission: {
+                    edit: "deny",
+                    bash: "deny",
+                },
+                prompt: `You are STRATEGOS - the Strategic General. You plan complex tasks through strategic interviewing and deep analysis.
+
+=== STRATEGOS WORKFLOW ===
+
+**PHASE 1: INTERVIEW (Critical!)**
+
+When the request is complex or unclear (new feature, architecture change, large refactor):
+
+1. Call the **strategos-interviewer** subagent
+2. Let the interviewer ask clarification questions
+3. Analyze the answers
+4. Ask follow-up questions if needed (more interviewer calls)
+
+**ACTIVATE Interview Mode for:**
+- Unclear requirements
+- New features (>1 file affected)
+- Architecture decisions
+- Integration with external systems
+- Performance-critical changes
+
+**PHASE 2: DISCOVERY**
+
+When interview is complete (or for simple requests):
+1. Call ALL explore-* subagents PARALLEL
+2. Analyze: context + code + deps + tests
+3. Synthesize findings
+
+**PHASE 3: STRATEGIC PLANNING**
+
+Create a comprehensive, strategic plan:
+
+# 🎯 Strategos Strategic Plan
+
+## 🎯 Mission Statement
+[1-2 sentences: What is the goal?]
+
+## 📋 Requirements Analysis
+### Must Have (Critical)
+- [Req 1]
+- [Req 2]
+
+### Should Have (Important)
+- [Req 3]
+
+### Could Have (Nice-to-have)
+- [Req 4]
+
+## 🏗️ Architecture Decisions
+| Decision | Rationale | Alternatives Considered |
+|----------|-----------|------------------------|
+| [Decision 1] | [Reasoning] | [Alternative] |
+| [Decision 2] | [Reasoning] | [Alternative] |
+
+## 📊 Implementation Roadmap
+
+### Phase 1: Foundation ([Time estimate])
+**Goal**: [What is achieved here?]
+**Deliverables**:
+- [ ] [Concrete Deliverable 1]
+- [ ] [Concrete Deliverable 2]
+
+**Files to Modify**:
+- \`path/to/file.ts\` - [Why?]
+- \`path/to/config.js\` - [Why?]
+
+**Steps**:
+1. [Concrete step with technology]
+2. [Concrete step]
+3. [Concrete step]
+
+### Phase 2: Core Implementation ([Time estimate])
+...
+
+### Phase 3: Integration & Testing ([Time estimate])
+...
+
+## ⚠️ Risk Assessment
+| Risk | Probability | Impact | Mitigation Strategy |
+|------|------------|--------|-------------------|
+| [Risk 1] | High/Medium/Low | High/Medium/Low | [How to avoid?] |
+| [Risk 2] | ... | ... | ... |
+
+## ✅ Quality Gates
+- [ ] Gate 1: [Criterion]
+- [ ] Gate 2: [Criterion]
+- [ ] Gate 3: [Criterion]
+
+## 🔄 Rollback Strategy
+If something goes wrong:
+1. [Step 1]
+2. [Step 2]
+
+## 🧪 Validation Plan
+[How do we test success?]
+
+## 📝 Open Questions
+- [ ] [Question 1] → [Recommended solution]
+- [ ] [Question 2] → [Recommended solution]
+
+---
+**Planned by**: Strategos Engine 🎯
+**Based on**: Interview + Codebase Analysis
+**Confidence Level**: [High/Medium/Low]
+
+**NEXT STEP**: Copy this plan and run in **Build** mode.
+
+=== STRATEGOS PRINCIPLES ===
+
+🎯 **Strategic Vision**: Anticipate problems BEFORE they occur
+🎯 **Tactical Depth**: Think in phases, not just steps
+🎯 **Risk Awareness**: Every plan has risks - name them
+🎯 **Clarity through Interviewing**: No assumptions, only clear requirements
+
+=== WHEN TO USE STRATEGOS ===
+
+Use Strategos for:
+- ✅ New features (>1 file)
+- ✅ Architecture changes
+- ✅ Integrations (APIs, DBs, external services)
+- ✅ Performance optimizations
+- ✅ Security improvements
+- ✅ Refactors with far-reaching consequences
+
+Use ultraplan for:
+- ✅ Simple bugfixes
+- ✅ Small features (1-2 files)
+- ✅ Documentation
+- ✅ Adding tests
+
+=== TODO SYSTEM 🔒 ===
+
+IMPORTANT: Create TODOs for each phase!
+
+## 📋 My TODOs
+- [ ] Phase 1: [Description]
+- [ ] Phase 2: [Description]
+- [ ] Phase 3: [Description]
+
+⚠️ You cannot stop until all Todos are completed!
+
+=== ACTIVATION KEYWORDS ===
+
+Strategos is automatically activated by:
+- "strategos" in the prompt
+- "interview mode" in the prompt
+- "strategic plan" in the prompt
+- "complex task" in the prompt
+- "🎯" emoji
+
+Or select the strategos Agent manually (Tab → strategos)`,
+            });
+            // Log configured agent models
+            const configuredAgents = [
+                'explore-context', 'explore-code', 'explore-deps', 'explore-tests',
+                'review-plan', 'strategos-interviewer',
+                'enhancer', 'ultraplan', 'ask', 'strategos'
+            ];
+            const modelSummary = configuredAgents
+                .map(name => `${name}=${input.agent[name]?.model || 'default'}`)
+                .join(', ');
+            client.app.log({
+                service: "enhancer",
+                level: "info",
+                message: `🔥 Registered ${Object.keys(input.agent).length} agents: ${Object.keys(input.agent).join(", ")}`
+            }).catch(() => { });
+            client.app.log({
+                service: "enhancer",
+                level: "debug",
+                message: `Agent models configured: ${modelSummary}`
+            }).catch(() => { });
         },
         "tool.execute.before": async (input, output) => {
             if (input.tool === "task") {
                 const args = output.args;
-                if (args?.subagent_type === "explore-context") {
+                const validSubagents = [
+                    "explore-context", "explore-code", "explore-deps",
+                    "explore-tests", "review-plan", "strategos-interviewer"
+                ];
+                if (args?.subagent_type && validSubagents.includes(args.subagent_type)) {
                     client.app.log({
                         service: "enhancer",
                         level: "debug",
-                        message: "Calling explore-context subagent"
+                        message: `Calling ${args.subagent_type} subagent`
                     }).catch(() => { });
                 }
             }
         },
         "message.updated": async (input, output) => {
-            if (input.session?.agent?.name !== "enhancer") {
+            const agentName = input.session?.agent?.name;
+            if (!["enhancer", "ultraplan", "ask", "strategos"].includes(agentName)) {
                 return;
             }
             const content = input.message?.content || "";
             if (content.includes("Build mode") || content.includes("Build Mode")) {
                 return;
             }
-            if (content.includes("```markdown") || content.includes("# Task:")) {
+            if (content.includes("\`\`\`markdown") || content.includes("# Task:")) {
                 const hint = "\n\n---\n**Next Step**: Copy this entire block and run it in **Build** mode to execute the plan.";
                 if (output && typeof output === "object") {
                     output.content = content + hint;
                 }
             }
-        }
+        },
+        // === TODO ENFORCER: Message Completed Hook ===
+        // Extracts todos from agent responses and stores them
+        "message.completed": async (input, output) => {
+            const agentName = input.session?.agent?.name;
+            const sessionId = input.session?.id || input.sessionId || 'default';
+            // Only for Todo-enforced agents
+            if (!TODO_ENFORCER_AGENTS.includes(agentName)) {
+                return;
+            }
+            try {
+                const content = input.message?.content || output?.content || "";
+                // Extract todos
+                const result = TodoParser.extractTodos(content);
+                if (result.hasTodos) {
+                    // Store todos
+                    todoStore.set(sessionId, result.todos, agentName);
+                    // Log for debugging
+                    await client.app.log({
+                        service: "enhancer",
+                        level: "info",
+                        message: `📝 Todos detected in ${agentName}: ${result.completedCount}/${result.todos.length} completed`
+                    }).catch(() => { });
+                    // Add visual todo block if pending todos exist
+                    if (result.pendingCount > 0) {
+                        const todoBlock = `\n\n${TODO_MARKER_START}\n## 📋 Open Todos (${result.pendingCount}/${result.todos.length})\n\n${TodoParser.formatTodoList(result.todos, true)}\n\n⚠️ **Note**: You cannot stop until all Todos are completed!\n${TODO_MARKER_END}`;
+                        if (output && typeof output === "object") {
+                            output.content = content + todoBlock;
+                        }
+                    }
+                }
+            }
+            catch (error) {
+                await client.app.log({
+                    service: "enhancer",
+                    level: "warn",
+                    message: `Error in todo extraction: ${error}`
+                }).catch(() => { });
+            }
+        },
+        // === TODO ENFORCER: Stop Requested Hook ===
+        // Prevents stop when todos are open
+        "stop.requested": async (input, output) => {
+            const agentName = input.session?.agent?.name;
+            const sessionId = input.session?.id || input.sessionId || 'default';
+            // Only for Todo-enforced agents
+            if (!TODO_ENFORCER_AGENTS.includes(agentName)) {
+                return;
+            }
+            try {
+                const unfinishedCount = todoStore.countUnfinished(sessionId);
+                if (unfinishedCount > 0) {
+                    const pendingTodos = todoStore.getAllPending(sessionId);
+                    // DENY STOP
+                    output.continue = true;
+                    output.reason = `🛑 **STOP DENIED** - ${unfinishedCount} TODO(S) STILL OPEN!\n\nYou cannot end the session until all tasks are completed. Please work on the remaining todos:\n\n${pendingTodos.map((t, i) => `${i + 1}. [ ] ${t.description}`).join('\n')}\n\n---\n**FORCED CONTINUATION MODE**\nThe agent must now complete the open todos.`;
+                    await client.app.log({
+                        service: "enhancer",
+                        level: "warn",
+                        message: `🛑 Stop blocked for ${agentName}: ${unfinishedCount} todos open`
+                    }).catch(() => { });
+                }
+                else {
+                    // All todos completed - clean up store
+                    if (todoStore.has(sessionId)) {
+                        todoStore.delete(sessionId);
+                        await client.app.log({
+                            service: "enhancer",
+                            level: "info",
+                            message: `✅ All todos completed in ${agentName} - Store cleaned`
+                        }).catch(() => { });
+                    }
+                }
+            }
+            catch (error) {
+                await client.app.log({
+                    service: "enhancer",
+                    level: "error",
+                    message: `Error in stop.requested hook: ${error}`
+                }).catch(() => { });
+            }
+        },
+        // === TODO ENFORCER: Session Start Hook ===
+        // Cleans old todos on new session
+        "session.start": async (input, output) => {
+            const agentName = input.agent?.name;
+            if (TODO_ENFORCER_AGENTS.includes(agentName)) {
+                // Cleanup old sessions
+                todoStore.cleanup();
+                await client.app.log({
+                    service: "enhancer",
+                    level: "debug",
+                    message: `🔄 New ${agentName} session started - Todo Store cleaned`
+                }).catch(() => { });
+            }
+        },
+        // === TODO ENFORCER: Session End Hook ===
+        "session.end": async (input, output) => {
+            const agentName = input.agent?.name;
+            const sessionId = input.session?.id || input.sessionId || 'default';
+            if (TODO_ENFORCER_AGENTS.includes(agentName)) {
+                // Check if open todos exist
+                if (todoStore.has(sessionId)) {
+                    const stats = todoStore.get(sessionId);
+                    const unfinished = stats?.todos.filter(t => t.status !== 'completed').length || 0;
+                    if (unfinished > 0) {
+                        await client.app.log({
+                            service: "enhancer",
+                            level: "warn",
+                            message: `⚠️ Session ended with ${unfinished} open todos`
+                        }).catch(() => { });
+                    }
+                    // Clean up
+                    todoStore.delete(sessionId);
+                }
+            }
+        },
+        // === STRATEGOS MODE ACTIVATION ===
+        // Auto-switch to strategos on keywords
+        "user.prompt.submitted": async (input, output) => {
+            const prompt = input.prompt || "";
+            const currentAgent = input.agent?.name;
+            const lowerPrompt = prompt.toLowerCase();
+            // Check for Strategos keywords
+            const detectedKeyword = STRATEGOS_KEYWORDS.find(kw => lowerPrompt.includes(kw.toLowerCase()));
+            if (detectedKeyword && currentAgent !== 'strategos') {
+                // Switch to strategos Agent
+                output.agent = 'strategos';
+                // Add hint
+                output.hint = `🎯 **Strategos Mode activated** (Keyword: "${detectedKeyword}")\n\nI will interview you first to clarify requirements, then create a strategic plan.`;
+                await client.app.log({
+                    service: "enhancer",
+                    level: "info",
+                    message: `🎯 Strategos Mode activated by keyword: "${detectedKeyword}"`
+                }).catch(() => { });
+            }
+            // Alternative: Interview Mode hint for ultraplan
+            if (lowerPrompt.includes('interview me') && currentAgent === 'ultraplan') {
+                output.hint = `💡 **Tip**: For complex tasks with interview, use the **strategos** Agent (Tab → strategos)`;
+            }
+        },
     };
 };
 export default EnhancerPlugin;