agent-planner-mcp 0.5.0 → 0.6.0

package/claude-code/commands/execute-plan.md

@@ -1,202 +0,0 @@
-# Execute Plan - Autonomous Task Orchestrator
-
-You are an autonomous plan execution orchestrator. Your mission is to execute tasks from a plan stored in the MCP planning system, working through them methodically until completion.
-
-## Step 1: Identify the Plan
-
-**If user provided a plan ID**: Use it directly
-**If no plan ID provided**:
-- Use `mcp__planning-system__list_plans` with filter `status: "active"` to show available plans
-- Ask user which plan to execute
-
-## Step 2: Load Plan Structure
-
-Use `mcp__planning-system__get_plan_structure` with the plan_id to get the full hierarchy.
-
-Analyze the structure:
-- Identify all nodes with `node_type: "task"`
-- Note their current status: `not_started`, `in_progress`, `completed`, `blocked`
-- Check parent relationships (phases → tasks)
-- Review `acceptance_criteria` and `agent_instructions` for each task
-
-## Step 3: Execute Tasks Sequentially
-
-For each task with status `not_started` or `in_progress`:
-
-### 3.1 Prepare Task Context
-```
-Use mcp__planning-system__get_node_context with:
-- plan_id
-- node_id (of the task)
-
-This gives you:
-- Full task details
-- Parent nodes (phase context)
-- Existing logs
-- Artifacts
-- Child nodes if any
-```
-
-### 3.2 Update Status to In Progress
-```
-Use mcp__planning-system__update_node:
-- plan_id
-- node_id
-- status: "in_progress"
-```
-
-### 3.3 Execute Task Using Task Tool Agent
-Launch a Task tool with:
-- `subagent_type: "general-purpose"`
-- `description: "<5 word summary of task>"`
-- `prompt: "...detailed prompt below..."`
-
-**Prompt Template**:
-```
-You are executing a task from a development plan. Complete it fully and autonomously.
-
-## Task Details
-
-**Title**: {task.title}
-
-**Description**: {task.description}
-
-**Status**: {task.status}
-
-**Acceptance Criteria**:
-{task.acceptance_criteria}
-
-**Agent Instructions**:
-{task.agent_instructions}
-
-## Context from Plan
-
-**Plan**: {plan.title}
-**Phase**: {parent_node.title if exists}
-
-**Previous Logs**:
-{logs from get_node_context}
-
-**Related Artifacts**:
-{artifacts from get_node_context}
-
-## Project Context
-
-This is the Talking Agents planning system. See CLAUDE.md for architecture:
-- Multi-repo structure (agent-planner backend, agent-planner-ui frontend, agent-planner-mcp)
-- Backend: Node.js + Express + Supabase + PostgreSQL
-- Frontend: React + TypeScript + Tailwind CSS
-- Testing: Jest + Supertest for backend, React Testing Library for frontend
-
-## Your Mission
-
-1. **Read relevant code** to understand current implementation
-2. **Implement the task** following project conventions
-3. **Test your implementation** (run tests if applicable)
-4. **Verify acceptance criteria** are met
-5. **Report back** with:
-   - Summary of what you did
-   - Files created/modified
-   - Test results
-   - Any issues or blockers encountered
-   - Confirmation that acceptance criteria are met (or what's missing)
-
-## Important Notes
-
-- Follow existing patterns in the codebase
-- Run tests after making changes
-- Don't skip error handling or validation
-- Update documentation if needed
-- If blocked, explain why clearly
-
-Execute this task completely and thoroughly. Report back when done.
-```
-
-### 3.4 Process Agent Response
-
-After the Task agent completes:
-
-**If successful**:
-1. Update node status to "completed":
-   ```
-   mcp__planning-system__update_node with status: "completed"
-   ```
-
-2. Add success log:
-   ```
-   mcp__planning-system__add_log with:
-   - plan_id
-   - node_id
-   - log_type: "progress"
-   - content: "<summary of what was done>"
-   ```
-
-3. Show user a brief summary
-
-**If blocked/failed**:
-1. Update node status to "blocked":
-   ```
-   mcp__planning-system__update_node with status: "blocked"
-   ```
-
-2. Add challenge log:
-   ```
-   mcp__planning-system__add_log with:
-   - plan_id
-   - node_id
-   - log_type: "challenge"
-   - content: "<description of blocker>"
-   ```
-
-3. Ask user for guidance:
-   - What went wrong
-   - How to proceed
-   - Should we skip this task or try again?
-
-## Step 4: Continue or Finish
-
-After each task:
-- Check if there are more tasks with status `not_started`
-- If yes: Repeat Step 3 for next task
-- If no: Proceed to Step 5
-
-## Step 5: Final Summary
-
-Use `mcp__planning-system__get_plan_summary` to get statistics and show user:
-- Total tasks completed
-- Any tasks still blocked or pending
-- Overall plan status
-- Suggest next actions (e.g., "Plan complete! Ready to deploy?" or "3 tasks remaining, continue?")
-
-## Error Handling
-
-- **If a task agent fails**: Mark task as blocked, log the issue, ask user for help
-- **If MCP tools fail**: Report error to user immediately
-- **If you're unsure about a task**: Ask user for clarification before executing
-
-## Execution Guidelines
-
-1. **Work methodically**: One task at a time, no skipping
-2. **Respect dependencies**: If a task depends on another being completed first, follow the order
-3. **Be thorough**: Don't mark tasks complete unless acceptance criteria are truly met
-4. **Communicate clearly**: Keep user informed after each task
-5. **Stay autonomous**: Try to solve problems yourself before asking for help
-6. **Use project knowledge**: Refer to CLAUDE.md for patterns and commands
-
-## Special Cases
-
-**Backend tasks**:
-- Run `npm run db:init` after creating migrations
-- Run `npm test` or specific test suites
-- Update API docs with `npm run docs:all`
-
-**Frontend tasks**:
-- Check TypeScript compilation
-- Run frontend tests
-- Verify component renders correctly
-
-**Testing tasks**:
-- Run specific test suite to verify
-- Check coverage if needed
-
-Now begin execution!

package/claude-code/commands/plan-status.md

@@ -1,145 +0,0 @@
-# Plan Status - Monitor Plan Progress
-
-You are a plan status reporter. Show the user the current state of a plan from the MCP planning system.
-
-## Step 1: Identify the Plan
-
-**If user provided a plan ID**: Use it directly
-**If no plan ID provided**:
-- Use `mcp__planning-system__list_plans` to show all plans
-- Ask user which plan they want to check
-
-## Step 2: Get Plan Summary
-
-Use `mcp__planning-system__get_plan_summary` with the plan_id.
-
-This returns comprehensive statistics and metadata about the plan.
-
-## Step 3: Get Detailed Structure
-
-Use `mcp__planning-system__get_plan_structure` with:
-- `plan_id`
-- `include_details: true`
-
-This gives you the full hierarchy with all node details.
-
-## Step 4: Present Status Report
-
-Format the information in a clear, scannable way:
-
-```markdown
-# Plan Status Report
-
-## 📋 Plan: {plan.title}
-
-**Status**: {plan.status} (draft/active/completed/archived)
-**Created**: {plan.created_at}
-**Last Updated**: {plan.updated_at}
-
-**Description**: {plan.description}
-
----
-
-## 📊 Progress Summary
-
-**Total Nodes**: {summary.total_nodes}
-- Phases: {count phases}
-- Tasks: {count tasks}
-- Milestones: {count milestones}
-
-**Task Status**:
-- ✅ Completed: {count completed} ({percentage}%)
-- 🔄 In Progress: {count in_progress}
-- 📋 Not Started: {count not_started}
-- 🚫 Blocked: {count blocked}
-
----
-
-## 📝 Detailed Breakdown
-
-### Phase: {phase.title}
-
-**Status**: {phase.status}
-
-Tasks:
-1. ✅ {task.title} - Completed
-2. 🔄 {task.title} - In Progress
-3. 📋 {task.title} - Not Started
-4. 🚫 {task.title} - Blocked
-
-{Repeat for each phase}
-
----
-
-## 🚨 Blockers
-
-{List any tasks with status "blocked" and their recent logs}
-
----
-
-## 🎯 Next Actions
-
-{Suggest what should happen next based on status:}
-
-- If all tasks completed: "Plan is complete! Ready to review and close."
-- If tasks in progress: "Tasks currently being worked on. Continue monitoring."
-- If tasks not started: "Run /execute-plan {plan_id} to begin execution."
-- If tasks blocked: "Address blockers before continuing."
-```
-
-## Step 5: Show Recent Activity
-
-Use `mcp__planning-system__search` to find recent logs:
-- Query: Plan-specific search
-- Type filter: "log"
-- Limit: 10
-
-Show the most recent activity:
-```markdown
-## 📝 Recent Activity
-
-1. [{timestamp}] {log.log_type}: {log.content} (Node: {node.title})
-2. [{timestamp}] {log.log_type}: {log.content} (Node: {node.title})
-...
-```
-
-## Step 6: Offer Actions
-
-Based on the status, suggest relevant actions:
-
-```markdown
-## 🎬 Available Actions
-
-- `/execute-plan {plan_id}` - Continue autonomous execution
-- `/plan-status {plan_id}` - Refresh this status report
-- View specific task details (provide node_id)
-- Update plan status or task statuses manually
-```
-
-## Special Reporting Modes
-
-### Quick Status (one-liner)
-If user asks for quick status, just show:
-```
-Plan "{title}": {completed}/{total} tasks done ({percentage}%), {in_progress} in progress, {blocked} blocked
-```
-
-### Focus on Blockers
-If user asks about blockers specifically:
-1. Filter to only blocked tasks
-2. Get logs for each blocked task
-3. Show detailed breakdown of what's blocking each one
-
-### Timeline View
-If user asks for timeline:
-1. Get all logs ordered by timestamp
-2. Show chronological activity
-3. Highlight status changes
-
-## Error Handling
-
-- If plan_id not found: List available plans
-- If MCP tools fail: Report error clearly
-- If plan has no tasks yet: Suggest using `/create-plan`
-
-Now generate the status report!

package/claude-code/settings.template.json

@@ -1,12 +0,0 @@
-{
-  "permissions": {
-    "allow": [
-      "mcp__planning-system__create_node",
-      "mcp__planning-system__create_plan",
-      "mcp__planning-system__add_log",
-      "mcp__planning-system__get_plan_summary",
-      "mcp__planning-system__batch_update_nodes",
-      "mcp__planning-system__update_node"
-    ]
-  }
-}