菜单
Create hyper-detailed task specifications with embedded context, eliminating implementation lookups. This skill should be used for breaking features into implementable tasks with complete architecture references.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/adolfoaranaes12/BMAD-ENHANCED --skill create-task-spec命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Comprehensive brownfield architecture analysis for existing codebases. Discovers structure, identifies patterns, assesses quality, calculates production readiness, and provides actionable recommendations. Use when analyzing existing codebases to understand architecture, assess quality, or prepare for modernization.
Systematically consume QA quality gate outputs and apply prioritized fixes for high-severity issues, NFR failures, and test coverage gaps. Use when quality review identifies issues requiring remediation.
Peer review architecture for quality, risks, and optimization opportunities. Analyzes scalability bottlenecks, security vulnerabilities, performance optimization, technology fit, and provides prioritized recommendations. Use when reviewing proposed architecture documents for quality assurance, risk identification, or architectural decision validation.
基于 SOC 职业分类
正在显示 SKILL.md
| name | create-task-spec |
| description | Create hyper-detailed task specifications with embedded context, eliminating implementation lookups. This skill should be used for breaking features into implementable tasks with complete architecture references. |
| acceptance | [{"task_file_created":"Task specification file created in configured location"},{"context_embedded":"All technical context embedded with source references"},{"tasks_defined":"3-15 sequential tasks with validation checkpoints"},{"user_approved":"User reviewed and approved specification"}] |
| inputs | {"requirement":{"type":"string","required":true,"description":"Feature or requirement to implement"},"priority":{"type":"string","required":false,"description":"Priority level (P0/P1/P2/P3)","default":"P2"}} |
| outputs | {"task_file":{"type":"string","description":"Path to created task specification file"},"task_id":{"type":"string","description":"Generated task ID (e.g., task-006)"},"task_count":{"type":"number","description":"Number of tasks in breakdown"},"status":{"type":"string","description":"Task status (Draft or Approved)"}} |
| telemetry | {"emit":"skill.create-task-spec.completed","track":["task_id","task_count","subtask_count","priority","status","duration_ms"]} |
Create hyper-detailed task specifications that embed ALL context needed for implementation, eliminating the need for developers to search for information during coding.
Core Capabilities:
BMAD Pattern (Key Innovation):
.claude/config.yaml) with task settings.claude/templates/task-spec.md)Action: Load configuration and verify prerequisites.
Execute:
python .claude/skills/bmad-commands/scripts/read_file.py \
--path .claude/config.yaml \
--output json
Extract: documentation.architecture, development.alwaysLoadFiles, development.taskLocation, templates.taskSpec
Verify: Task location exists, template exists, determine next task ID
Halt if: Config missing, template missing, cannot determine task ID
See: references/templates.md for configuration format and references/requirements-gathering-guide.md for requirement clarification
Action: Clarify requirement, acceptance criteria, and priority.
Ask user:
What needs to be implemented?
What defines "done"?
Priority and complexity?
Confirm understanding:
Halt if:
See: references/requirements-gathering-guide.md for detailed gathering techniques
Action: Load architecture documentation and previous task learnings.
Load files:
Always-required files (coding standards):
python .claude/skills/bmad-commands/scripts/read_file.py \
--path {always_load_file} \
--output json
Architecture documentation:
documentation.architecture filesPrevious task insights:
taskLocationContext categories: Tech stack, project structure, coding standards, testing strategy, data models, API specs, component specs, patterns
CRITICAL RULES: ONLY use information from source documents, NEVER invent details, ALWAYS cite sources, note when info missing
See: references/context-extraction-guide.md for detailed extraction strategies and references/templates.md for context format examples
Action: Identify affected components and extract specific technical details.
Identify components:
Extract details: Data models (schemas, validation, relationships), API specs (endpoints, request/response, auth, errors), component specs (location, props, state, styling), file locations (exact paths), constraints (performance, security, reliability, testing)
All extractions must cite: [Source: filename#section]
See: references/context-extraction-guide.md for analysis patterns and references/templates.md for extraction format examples
Action: Break requirement into 3-15 sequential tasks with subtasks.
Task breakdown:
Break into 3-15 tasks:
Create subtasks for each task:
Structure in implementation order:
Link to acceptance criteria: Example: "Task 1: Create user model (AC: 1, 2)"
Halt if: Cannot break down coherently, >15 tasks (too complex), <3 tasks (too simple)
See: references/task-breakdown-guide.md for detailed strategies and references/templates.md for task structure examples
Action: Generate task file from template with all context embedded.
Load template:
python .claude/skills/bmad-commands/scripts/read_file.py \
--path .claude/templates/task-spec.md \
--output json
Replace placeholders: task_id, title, date, priority, status, user_story, acceptance_criteria, context, tasks, validation
Context must include: Previous insights, data models, API specs, component specs, file locations, testing requirements, constraints - ALL with [Source: filename#section] references
See: references/templates.md for placeholder examples and formats
Action: Verify all required elements are present and properly formatted.
Validation checklist:
Required Sections:
Context Completeness:
Task Quality:
Halt if:
See: references/validation-approval-guide.md for validation details
Action: Save task file and present summary for approval.
Save file:
python .claude/skills/bmad-commands/scripts/write_file.py \
--path {taskLocation}/{task_id}.md \
--content "{task_spec}" \
--output json
Present summary: Show task ID, objective, acceptance criteria, task breakdown, embedded context sources, and request approval.
Update status if approved: Change from "Draft" to "Approved"
See: references/templates.md for complete summary format and references/validation-approval-guide.md for approval workflow
Return structured output with task file path, task ID, task count, status, and telemetry metrics.
See: references/templates.md for complete output format
See: references/task-breakdown-guide.md for detailed practices
Use this skill when:
Always use before:
Do not use for:
references/requirements-gathering-guide.md - Clarifying requirements, acceptance criteria, user storiesreferences/context-extraction-guide.md - Loading architecture, extracting technical details, sourcingreferences/task-breakdown-guide.md - Creating sequential tasks, granularity, validation checkpointsreferences/validation-approval-guide.md - Completeness validation, saving, approval workflowreferences/templates.md - Task spec template, output format, placeholder examples, configuration formatInvoked with requirement input (feature description) and optional priority (P0-P3)
Part of BMAD Enhanced Planning Suite