@e0ipso/ai-task-manager 1.26.11 → 1.27.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@e0ipso/ai-task-manager",
-  "version": "1.26.11",
+  "version": "1.27.1",
   "description": "Task management for AI coding assistants",
   "main": "dist/index.js",
   "bin": {

package/templates/ai-task-manager/config/hooks/POST_EXECUTION.md

@@ -0,0 +1,19 @@
+# POST_EXECUTION Hook
+
+## Validation Gates
+
+Before marking the blueprint as complete, verify:
+
+- [ ] All linting rules must pass without errors. If no linter is configured, skip this step
+- [ ] All tests must pass successfully. If no test suite is configured, skip this step
+- [ ] Verify all tasks in the plan have `status: "completed"` in their frontmatter
+- [ ] Verify that the AGENTS.md documentation or related documentes are still correct after this plan execution
+
+## Failure Behavior
+
+If any validation gate fails:
+
+- **Halt execution immediately** - do not proceed to summary generation or archival
+- **Leave plan in `plans/` directory** for debugging and correction
+- **Document the failure** in the plan file with details about which gate failed
+- **Provide actionable next steps** for resolving the failure

package/templates/ai-task-manager/config/hooks/POST_TASK_GENERATION_ALL.md

@@ -1,162 +1,41 @@
 # POST_TASK_GENERATION_ALL Hook
 
-After all tasks have been generated use your internal Todo task to execute:
+After all tasks have been generated, perform these two steps:
 
-- [ ] Complexity Analysis & Refinement
-- [ ] Update plan document with Blueprint information
+## 1. Review Task Complexity
 
-## Complexity Analysis & Refinement
+For each generated task, do a quick sanity check:
 
-### Complexity Scoring Matrix (1-10 scale)
+- **Too complex?** If a task spans 3+ technologies or requires 3+ skills, split it.
+- **Too vague?** If acceptance criteria are unclear, sharpen them.
+- **Too trivial?** If two tasks could be one without adding complexity, merge them.
 
-| Dimension | 1-2 | 3-4 | 5-6 | 7-8 | 9-10 |
-|-----------|-----|-----|-----|-----|------|
-| **Technical** | Basic ops | Single tech | 2-3 techs | Multiple complex | Cutting-edge |
-| **Decision** | No decisions | 1-2 minor | Trade-offs | Interdependent | Novel solutions |
-| **Integration** | Single file | 2-3 files | Multi-module | Many systems | 15+ services |
-| **Scope** | Atomic action | Small feature | Complete feature | Major feature | Entire subsystem |
-| **Uncertainty** | Crystal clear | Minor ambiguity | Some clarification | Research required | Experimental |
+Target: every task should be completable with 1-2 skills and have clear acceptance criteria.
 
-**Composite Score** = MAX(Technical×1.0, Decision×0.9, Integration×0.8, Scope×0.7, Uncertainty×1.1)
+## 2. Update Plan with Blueprint
 
-### Decomposition Rules
+After finalizing tasks, append to the plan document:
 
-**Thresholds:**
-- Composite ≥6: Consider decomposition
-- Any dimension ≥8: Mandatory decomposition
-- Multiple dimensions ≥6: High priority
+### Dependency Diagram
 
-**Patterns by Driver:**
-| Driver | Pattern | Strategy |
-|--------|---------|----------|
-| Technical | Technology Layering | Split by tech boundaries |
-| Decision | Decision-Implementation | Separate analysis from execution |
-| Integration | Integration Isolation | Core feature + integrations |
-| Scope | Functional | Split by user workflows |
-| Uncertainty | Research-Implementation | Prototype then implement |
-
-### AIDVR Process
-1. **Assess**: Confirm need (score >5)
-2. **Identify**: Find natural breakpoints
-3. **Decompose**: Apply pattern
-4. **Validate**: Check criteria
-5. **Reconstruct**: Update dependencies
-
-### Safety Controls
-
-**Limits:**
-- Max 3 decomposition rounds per task
-- No decomposition if score ≤3
-- Min 2-hour work per subtask
-- Stop if complexity not decreasing
-
-**Stop Conditions:**
-1. Atomic boundary reached
-2. <1 skill per subtask
-3. Overhead > benefit
-4. Resource fragmentation
-5. <2 hour granularity
-6. 3 iterations reached
-7. Score ≤3
-8. No complexity reduction
-
-### Validation Checklist
-
-**Per Subtask:**
-- [ ] Complexity ≤5 (target ≤4)
-- [ ] Skills ≤2
-- [ ] Clear acceptance criteria
-- [ ] Dependencies acyclic
-- [ ] Score ≥2 (min viability)
-
-**Decomposition Quality:**
-- [ ] Coverage: All requirements included
-- [ ] Coherence: Single purpose per task
-- [ ] Consistency: Lower scores than parent
-- [ ] Coordination: Integration plan exists
-
-### Error Handling
-
-| Issue | Detection | Resolution |
-|-------|-----------|------------|
-| Infinite loop | No reduction after 2 rounds | Stop, mark needs-clarification |
-| Circular dependency | Task reaches itself | Break shortest cycle |
-| Over-decomposition | Score <2 or <2hrs | Merge with sibling |
-| Orphaned task | Missing dependencies | Create or remove dependency |
-| Scope creep | Subtasks > original | Remove excess scope |
-| Skill conflicts | >2 skills needed | Decompose by skill boundary |
-| High complexity | >7 after max rounds | Escalate with documentation |
-
-### Documentation Format
-
-In `complexity_notes`:
-```
-Original: 8.2 → Round 1 → [4.1, 3.8, 4.5] → PASSED
-```
-
-**Quality Gates Before Proceeding:**
-- All tasks ≤5 complexity or escalated
-- No unresolved errors
-- Iterations documented
-- Validation complete
-
-## Update plan document with Blueprint information
-
-After creating all tasks with their dependencies, update the original plan document with two critical sections: a task dependency visualization and a phase-based execution blueprint.
-
-### Section 1: Dependency Visualization
-
-If any tasks have dependencies, create a Mermaid diagram showing the dependency graph:
+If tasks have dependencies, add a Mermaid graph:
 
 ```mermaid
 graph TD
-    001[Task 001: Database Schema] --> 002[Task 002: API Endpoints]
-    001 --> 003[Task 003: Data Models]
-    002 --> 004[Task 004: Frontend Integration]
-    003 --> 004
+    001[Task 001: Description] --> 002[Task 002: Description]
 ```
 
-Note: Ensure the graph is acyclic (no circular dependencies).
-
-### Section 2: Phase-Based Execution Blueprint
-
-#### Core Concept
-The execution blueprint organizes tasks into sequential phases where:
-- **Within a phase**: All tasks execute in parallel
-- **Between phases**: Execution is strictly sequential
-- **Phase progression**: Requires all tasks in current phase to complete AND validation gates to pass
-
-#### Phase Definition Rules
-1. **Phase 1**: Contains all tasks with zero dependencies
-2. **Phase N**: Contains tasks whose dependencies are ALL satisfied by tasks in phases 1 through N-1
-3. **Parallelism Priority**: Maximize the number of tasks that can run simultaneously in each phase
-4. **Completeness**: Every task must be assigned to exactly one phase
-
-#### Blueprint Structure
-
-Use the template in .ai/task-manager/config/templates/BLUEPRINT_TEMPLATE.md for the execution blueprint structure.
-
-### Validation Requirements
-
-#### Phase Transition Rules
-1. All tasks in the current phase must have status: "completed"
-2. All validation gates defined in `/config/hooks/POST_PHASE.md` for the current phase must pass
-3. No task in a future phase can begin until these conditions are met
+Verify there are no circular dependencies.
 
-#### Blueprint Verification
-Before finalizing, ensure:
-- [ ] Every task appears in exactly one phase
-- [ ] No task appears in a phase before all its dependencies
-- [ ] Phase 1 contains only tasks with no dependencies
-- [ ] Each phase maximizes parallel execution opportunities
-- [ ] All phases reference their validation gates
-- [ ] The execution summary accurately reflects the blueprint
+### Execution Phases
 
-### Important Notes
+Group tasks into phases:
+- **Phase 1**: Tasks with no dependencies (run in parallel)
+- **Phase N**: Tasks whose dependencies are all in earlier phases
 
-### Error Handling
+Use the template in `.ai/task-manager/config/templates/BLUEPRINT_TEMPLATE.md` for structure.
 
-If dependency analysis reveals issues:
-- **Circular dependencies**: Document the cycle and mark affected tasks for review
-- **Orphaned tasks**: Tasks that cannot be scheduled due to missing dependencies
-- **Ambiguous dependencies**: Note assumptions made and flag for clarification
+Before finalizing, verify:
+- Every task is in exactly one phase
+- No task runs before its dependencies complete
+- Phase 1 has only zero-dependency tasks

package/templates/ai-task-manager/config/hooks/PRE_TASK_ASSIGNMENT.md

@@ -11,19 +11,13 @@ This hook executes before task assignment to determine the most appropriate agen
     - Select the most appropriate sub-agent (if any are available). If no sub-agent is appropriate, use the general-purpose one.
     - Consider task-specific requirements from the task document
 
-<details>
-<summary>
-[IMPORTANT] Only for assistants that support assistant skills: Claude
-</summary>
-
-Analyze the set of tasks skills in order to engage any relevant assistant skills as necessary (either global or project skills).
-</details>
+[IMPORTANT] Analyze the set of tasks skills in order to engage any relevant assistant skills as necessary (either global
+or project skills).
 
 
 ## Available Sub-Agents
-Analyze the sub-agents available in your current assistant's agents directory. If none are available
-or the available ones do not match the task's requirements, then use a generic
-agent.
+Analyze the sub-agents available in your current assistant's agents directory. If none are available or the available
+ones do not match the task's requirements, then use a generic agent.
 
 ## Matching Criteria
 Select agents based on:
@@ -38,27 +32,7 @@ Read task skills and select appropriate task-specific agent:
 
 ```bash
 # Extract skills from task frontmatter
-TASK_SKILLS=$(awk '
-    /^---$/ { if (++delim == 2) exit }
-    /^skills:/ {
-        in_skills = 1
-        # Check if skills are on the same line
-        if (match($0, /\[.*\]/)) {
-            gsub(/^skills:[ \t]*\[/, "")
-            gsub(/\].*$/, "")
-            gsub(/[ \t]/, "")
-            print
-            in_skills = 0
-        }
-        next
-    }
-    in_skills && /^[^ ]/ { in_skills = 0 }
-    in_skills && /^[ \t]*-/ {
-        gsub(/^[ \t]*-[ \t]*/, "")
-        gsub(/^"/, ""); gsub(/"$/, "")
-        print
-    }
-' "$TASK_FILE" | tr ',' '\n' | sed 's/^[ \t]*//;s/[ \t]*$//' | grep -v '^$')
+TASK_SKILLS=$(node "$root/config/scripts/extract-task-skills.cjs" "$TASK_FILE")
 
 echo "Task skills required: $TASK_SKILLS"
 
@@ -75,4 +49,4 @@ done
 if [ "$AGENT_FOUND" = false ]; then
     echo "Using general-purpose agent for task execution"
 fi
-```
+```

package/templates/ai-task-manager/config/scripts/extract-task-skills.cjs

@@ -0,0 +1,84 @@
+#!/usr/bin/env node
+
+/**
+ * Extract skills from a task file's YAML frontmatter.
+ *
+ * Usage: node extract-task-skills.cjs <task-file>
+ * Output: One skill per line, trimmed, empty lines removed.
+ *
+ * Supports both inline array and block list formats:
+ *   skills: [skill-a, skill-b]
+ *   skills:
+ *     - skill-a
+ *     - "skill-b"
+ */
+
+'use strict';
+
+const fs = require('fs');
+const path = require('path');
+
+const taskFile = process.argv[2];
+if (!taskFile) {
+  console.error('Usage: node extract-task-skills.cjs <task-file>');
+  process.exit(1);
+}
+
+const content = fs.readFileSync(path.resolve(taskFile), 'utf8');
+const lines = content.split('\n');
+
+let inFrontmatter = false;
+let frontmatterClosed = false;
+let inSkills = false;
+const skills = [];
+
+for (const line of lines) {
+  if (frontmatterClosed) break;
+
+  if (line.trim() === '---') {
+    if (!inFrontmatter) {
+      inFrontmatter = true;
+      continue;
+    }
+    // Second delimiter – end of frontmatter
+    frontmatterClosed = true;
+    break;
+  }
+
+  if (!inFrontmatter) continue;
+
+  // Detect "skills:" key
+  if (/^skills\s*:/.test(line)) {
+    inSkills = true;
+
+    // Check for inline array: skills: [a, b, c]
+    const inlineMatch = line.match(/\[(.*)]/);
+    if (inlineMatch) {
+      inlineMatch[1]
+        .split(',')
+        .map(s => s.trim().replace(/^["']|["']$/g, ''))
+        .filter(Boolean)
+        .forEach(s => skills.push(s));
+      inSkills = false;
+    }
+    continue;
+  }
+
+  if (inSkills) {
+    // A non-indented, non-empty line ends the block list
+    if (/^\S/.test(line)) {
+      inSkills = false;
+      continue;
+    }
+    // List item: "  - skill-name"
+    const itemMatch = line.match(/^\s*-\s*(.*)/);
+    if (itemMatch) {
+      const value = itemMatch[1].trim().replace(/^["']|["']$/g, '');
+      if (value) skills.push(value);
+    }
+  }
+}
+
+if (skills.length > 0) {
+  console.log(skills.join('\n'));
+}

package/templates/assistant/commands/tasks/create-plan.md

@@ -10,7 +10,9 @@ ensuring you have all necessary context before proceeding. Use the plan-creator
 
 ---
 
-Ultrathink, think harder, and use tools.
+Use tools for the planning. You are encouraged to write your own specialized tools to research, analyze, and debug
+any work order from the user. You are not restricted to the stack of the current project to create your own
+specialized tools.
 
 ## Find the AI Task Manager root
 
@@ -34,7 +36,7 @@ const findRoot = (currentDir) => {
   }
 
   const parentDir = path.dirname(currentDir);
-  if (parentDir !== currentDir) {
+  if (parentDir.length < currentDir.length) {
     findRoot(parentDir);
   } else {
     process.exit(1);

package/templates/assistant/commands/tasks/execute-blueprint.md

@@ -55,7 +55,7 @@ const findRoot = (currentDir) => {
   }
 
   const parentDir = path.dirname(currentDir);
-  if (parentDir !== currentDir) {
+  if (parentDir.length < currentDir.length) {
     findRoot(parentDir);
   } else {
     process.exit(1);
@@ -123,6 +123,7 @@ Use your internal Todo task tool to track the execution of all phases, and the f
 - [ ] Execute $root/.ai/task-manager/config/hooks/PRE_PHASE.md hook before Phase 3.
 - [ ] Phase 3: Execute 1 task(s) in parallel.
 - [ ] Execute $root/.ai/task-manager/config/hooks/POST_PHASE.md hook after Phase 3.
+- [ ] Execute $root/.ai/task-manager/config/hooks/POST_EXECUTION.md hook after all phases complete.
 - [ ] Update the Plan 7 with execution summary using $root/.ai/task-manager/config/hooks/EXECUTION_SUMMARY_TEMPLATE.md.
 - [ ] Archive Plan 7.
 
@@ -198,9 +199,16 @@ This structured output enables automated workflow coordination and must be inclu
 
 Upon successful completion of all phases and validation gates, perform the following additional steps:
 
+- [ ] Post-Execution Validation
 - [ ] Execution Summary Generation
 - [ ] Plan Archival
 
+### 0. Post-Execution Validation
+
+Read and execute $root/.ai/task-manager/config/hooks/POST_EXECUTION.md
+
+If validation fails, halt execution. The plan remains in `plans/` for debugging.
+
 ### 1. Execution Summary Generation
 
 Append an execution summary section to the plan document with the format described in $root/.ai/task-manager/config/templates/[EXECUTION_SUMMARY_TEMPLATE.md

package/templates/assistant/commands/tasks/execute-task.md

@@ -30,7 +30,7 @@ const findRoot = (currentDir) => {
   }
 
   const parentDir = path.dirname(currentDir);
-  if (parentDir !== currentDir) {
+  if (parentDir.length < currentDir.length) {
     findRoot(parentDir);
   } else {
     process.exit(1);

package/templates/assistant/commands/tasks/fix-broken-tests.md

@@ -3,7 +3,11 @@ argument-hint: "[testCommand]"
 description: Fix the tests your task execution broke.
 ---
 
-Fix all failing tests in this repository. Ultrathink, think harder, and use tools.
+Fix all failing tests in this repository.
+
+Use tools for debugging. You are encouraged to write your own specialized tools to research, analyze, and debug
+failing tests. You are not restricted to the stack of the current project to create your own
+specialized tools.
 
 Execute this command to run the tests:
 

package/templates/assistant/commands/tasks/full-workflow.md

@@ -31,7 +31,7 @@ const findRoot = (currentDir) => {
   }
 
   const parentDir = path.dirname(currentDir);
-  if (parentDir !== currentDir) {
+  if (parentDir.length < currentDir.length) {
     findRoot(parentDir);
   } else {
     process.exit(1);
@@ -443,9 +443,16 @@ This structured output enables automated workflow coordination and must be inclu
 
 Upon successful completion of all phases and validation gates, perform the following additional steps:
 
+- [ ] Post-Execution Validation
 - [ ] Execution Summary Generation
 - [ ] Plan Archival
 
+#### Post-Execution Validation
+
+Read and execute $root/.ai/task-manager/config/hooks/POST_EXECUTION.md
+
+If validation fails, halt execution. The plan remains in `plans/` for debugging.
+
 #### Execution Summary Generation
 
 Append an execution summary section to the plan document with the format described in $root/.ai/task-manager/config/templates/EXECUTION_SUMMARY_TEMPLATE.md

package/templates/assistant/commands/tasks/generate-tasks.md

@@ -7,8 +7,6 @@ description: Generate tasks to implement the plan with the provided ID.
 
 ---
 
-Ultrathink, think harder, and use tools.
-
 ## Find the AI Task Manager root
 
 ```bash
@@ -31,7 +29,7 @@ const findRoot = (currentDir) => {
   }
 
   const parentDir = path.dirname(currentDir);
-  if (parentDir !== currentDir) {
+  if (parentDir.length < currentDir.length) {
     findRoot(parentDir);
   } else {
     process.exit(1);

package/templates/assistant/commands/tasks/refine-plan.md

@@ -11,7 +11,9 @@ user, and update the plan with the refinements. Use the plan-creator sub-agent f
 
 ---
 
-Ultrathink, think harder, and use tools.
+Use tools for the plan refinement. You are encouraged to write your own specialized tools to research, analyze, and
+debug any plan refinement. You are not restricted to the stack of the current project to create your own specialized
+tools.
 
 ## Find the AI Task Manager root
 
@@ -35,7 +37,7 @@ const findRoot = (currentDir) => {
   }
 
   const parentDir = path.dirname(currentDir);
-  if (parentDir !== currentDir) {
+  if (parentDir.length < currentDir.length) {
     findRoot(parentDir);
   } else {
     process.exit(1);