name: execute-task description: Execute approved task specifications sequentially with TDD, comprehensive testing, and validation. This skill should be used for implementing tasks from approved specs with full audit trail. acceptance:

task_loaded: "Task specification successfully loaded and validated"
status_updated: "Task status progressed from Approved → InProgress → Review"
all_complete: "All tasks and subtasks marked complete"
tests_passing: "All tests written and passing"
criteria_verified: "All acceptance criteria verified and documented"
record_complete: "Implementation Record fully populated" inputs: task_file: type: string required: true description: "Path to task specification file (e.g., .claude/tasks/task-006.md)" validation: "Must be existing file with status 'Approved'" auto_confirm: type: boolean required: false description: "Skip execution plan confirmation" default: false outputs: implementation_complete: type: boolean description: "Whether all tasks completed successfully" tasks_completed: type: number description: "Number of tasks completed" subtasks_completed: type: number description: "Number of subtasks completed" tests_passed: type: boolean description: "Whether all tests passed" total_tests: type: number description: "Total number of tests executed" files_modified: type: array description: "List of files created or modified" status: type: string description: "Final task status (should be 'Review')" telemetry: emit: "skill.execute-task.completed" track:
- task_file
- tasks_completed
- subtasks_completed
- tests_passed
- total_tests
- files_modified_count
- duration_ms
- halt_count

Execute Task Skill

Purpose

Execute approved task specifications sequentially using Test-Driven Development (TDD), with comprehensive validation at each step and full implementation audit trail.

Core Capabilities:

Sequential task/subtask execution with validation gates
Permission-controlled file modifications
Comprehensive testing and documentation
Implementation Record maintenance
Halt-on-error safeguards

BMAD Pattern (Key Innovation):

ALL context embedded in task spec by planning skill
Dev reads ONLY task spec + always-load files (coding standards)
No architecture lookup during implementation
Result: Focused execution, no context searching, no drift

Prerequisites

Task specification file with status "Approved"
Configuration file (.claude/config.yaml) with development settings
Always-load files (coding standards) available
Test framework configured (Jest, Pytest, etc.)

Workflow

Step 0: Load Configuration and Task Specification

Action: Load all required context and verify task readiness.

Execute:

python .claude/skills/bmad-commands/scripts/read_file.py \
  --path .claude/config.yaml \
  --output json

python .claude/skills/bmad-commands/scripts/read_file.py \
  --path {task_file} \
  --output json

Parse Response:

Extract development.alwaysLoadFiles from config
Extract task status, objective, acceptance criteria, tasks
Verify status is "Approved"

Validation:

Configuration file exists and loadable
Task file exists and loadable
Task status is "Approved" (halt if Draft/Review/Done)
Always-load files exist

Update Status:

Change task status from "Approved" to "InProgress"
Record start time in Implementation Record

See: references/configuration-guide.md for detailed configuration loading

Step 1: Review Task Context and Plan Execution

Action: Review embedded context and present execution plan to user.

Context Review:

Previous task insights
Data models and schemas
API specifications
Component specifications
File locations
Testing requirements
Technical constraints

Task Breakdown:

Count total tasks and subtasks
Identify current task (first unchecked)
Understand task dependencies

Present Plan: Display execution plan with task name, file, context loaded (task spec + coding standards), execution sequence (numbered tasks with subtask counts), total counts, confirmation prompt

See: references/templates.md#execution-plan-template for complete format

Wait for confirmation unless auto_confirm=true

Halt Conditions:

Context appears insufficient
Task breakdown unclear
User does not confirm

See: references/task-execution-guide.md for execution details

Step 2: Execute Current Task

Action: Execute each task and subtask sequentially with validation.

For each task in sequence:

Announce task: Display task name with acceptance criteria and subtask list
For each subtask: Implement (read context, create/modify files per coding standards) | If "Write tests": create and run tests (must pass) | If "Validate": run tests, linter, verify AC | Update checkbox to [x] only when complete | Record notes in Implementation Record (deviations, decisions, learnings)
After all subtasks: Run task validation (all tests, lint, AC coverage) | Update task checkbox [x] only when validated | Update Implementation Record (files, notes)
Move to next task

Halt Conditions:

3 consecutive implementation failures on same subtask
Ambiguous requirements discovered
Missing dependencies not documented
Regression test failures
User requests halt

See: references/task-execution-guide.md for detailed execution examples

Step 3: Final Validation and Documentation

Action: Run complete validation and finalize documentation.

Validation:

Run complete test suite:

python .claude/skills/bmad-commands/scripts/run_tests.py \
  --path . \
  --framework auto \
  --output json

Verify acceptance criteria:
- Review each AC from task spec
- Map AC to implementation and tests
- Confirm all ACs covered
Verify all checkboxes marked:
- Scan task spec for unchecked [ ] boxes
- Ensure all tasks and subtasks complete

Documentation: Update Implementation Record with agent model, completion notes (details, decisions, learnings), files modified (created/modified lists), testing results (unit/integration/regression test counts, coverage %, execution time)

See: references/templates.md#implementation-record-complete-template for complete format

Status Update:

Change status from "InProgress" to "Review"
DO NOT mark as "Done" (quality skill does that)

Present Summary: Display completion summary with task name, status (Review), what was implemented, all ACs met (with checkmarks), test results (counts, coverage, regression), files created/modified counts, quality review prompt

See: references/templates.md#completion-summary-template for complete format

See: references/validation-guide.md for validation details

Step 4: Handle Quality Review (Optional)

Action: Provide next steps based on user decision.

If user requests quality review: Confirm task marked "Review", provide next step (use quality review skill with task file)

If user approves without review: Confirm approval, provide next steps (commit changes, mark "Done", move to next task)

See: references/templates.md#step-4-handle-quality-review for complete messages

File Modification Permissions

CRITICAL PERMISSION BOUNDARIES:

YOU ARE AUTHORIZED TO:

✅ Update "Implementation Record" section of task file
✅ Update task/subtask checkboxes ([ ] to [x])
✅ Update task status line (Approved → InProgress → Review)
✅ Create, modify, delete implementation files (src/, tests/, etc.)
✅ Run commands (tests, linters, builds)

YOU ARE NOT AUTHORIZED TO:

❌ Modify "Objective" section of task file
❌ Modify "Acceptance Criteria" section of task file
❌ Modify "Context" section of task file
❌ Modify task/subtask descriptions (only checkboxes)
❌ Modify "Quality Review" section of task file
❌ Change task status to "Done" (only to "Review")

Enforcement: Only edit Implementation Record section and checkboxes/status

See: references/permissions-halts.md for detailed permission boundaries

Halt Conditions

Must halt execution and ask user when:

Consecutive Failures (default: 3)
- Same subtask fails 3 times in a row
- Present error and ask for guidance
Ambiguous Requirements
- Context insufficient to implement subtask
- Multiple valid interpretations
- Critical technical decision needed
Missing Dependencies
- Required library/service not documented
- External API credentials needed
- Database not accessible
Regression Failures
- Existing tests start failing
- Breaking change introduced
User Interruption
- User requests halt
- User asks question mid-execution

Halt Message Format: Display halt warning with reason (category), context (what was attempted), issue (specific problem), need from user (required info/decision), current progress (tasks/subtasks complete vs remaining), ready to resume condition

See: references/templates.md#halt-message-templates for all halt types

See: references/permissions-halts.md for halt handling details

Output

Return structured JSON output with implementation_complete (boolean), tasks_completed, subtasks_completed, tests_passed, total_tests, files_modified (array), status, telemetry (skill, task_file, metrics, duration_ms, halt_count)

See: references/templates.md#complete-json-output-format for full structure and examples (success, with halts, failed)

Error Handling

If any step fails:

1. Task File Not Found:

Error: "Task file not found"
Action: Verify file path

2. Task Status Not Approved:

Error: "Task must be Approved before execution"
Action: Check task status, update if needed

3. Test Failures:

Error: "X tests failing"
Action: Review failures, fix issues, re-run

4. Missing Dependencies:

Error: "Dependency X not found"
Action: Verify task spec includes dependency info

Best Practices

Trust the task spec (context embedded, don't search) | Sequential execution (complete current before next) | Test before checking (run tests before marking [x]) | Document as you go (update notes after each task) | Respect permissions (only Implementation Record + checkboxes) | Halt when appropriate (don't guess unclear requirements)

See: references/best-practices.md and references/templates.md#best-practices-with-examples for detailed guidance and examples

Routing Guidance

Use this skill when:

Executing an approved task specification
Need sequential, validated task execution
Want comprehensive testing and documentation
Need implementation audit trail

Always use after:

Task spec created and approved
Planning complete

Before:

Quality review
Pull request creation

Reference Files

Detailed documentation in references/:

templates.md: All output formats, examples, file templates, integration workflows, JSON structures, error templates, command-line usage, best practices with examples
configuration-guide.md: Loading config, always-load files, status management
task-execution-guide.md: Executing tasks/subtasks, validation gates, examples
validation-guide.md: Final validation, documentation, completion summary
implementation-record.md: Templates for Implementation Record section
permissions-halts.md: Permission boundaries, halt conditions, error handling
best-practices.md: Best practices, pitfalls to avoid, integration patterns

Part of BMAD Enhanced Implementation Suite

execute-task

$ 安裝