Token导航 LogoToken导航TokenDH.com
研究检索需要联网github未标认证来源可访问clear审计未展示

spec-creation规格创建

Agent Skill

spec-creation 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

220

周安装

9

GitHub Stars

公开资料未说明

下载量

71
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:spec-creation(规格创建)
来源仓库:https://github.com/davjdk/thermocalcbot
仓库路径:skills/spec-creation
安装命令:
npx skills add davjdk/thermocalcbot --skill "spec-creation"
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

AgentSkills.tonpx skills
npx skills add davjdk/thermocalcbot --skill "spec-creation"

简介

创建产品功能与技术实现规格文档。

  • 引导团队对齐需求边界与验收条件。
  • 适用于敏捷开发前期的规划阶段。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。
  • 应结合用户故事与原型进行细化。
  • spec-creation 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Spec Creation Skill

Create comprehensive feature specifications for the Web Portal project through iterative requirement refinement based on ISO/IEC/IEEE 29148:2018 quality criteria.

When to Use This Skill

  • User requests creating a technical specification (ТЗ, техническое задание)
  • User says "создай ТЗ", "напиши спецификацию", "создай spec"
  • Starting workflow step "а" - Создание ТЗ (per CLAUDE.md workflow)
  • User wants to document a feature before implementation

Output Location

Specifications are created in .ai/specs/ directory with naming pattern:

  • feature-{slug}.md for new features
  • bugfix-{slug}.md for bug fixes

Core Principles

Iterative Requirement Refinement

This skill uses multi-round validation with IEEE 29148:2018 quality criteria:

Critical: A specification is NOT complete until validation achieves ≥85% score.

Progressive Disclosure

  • Keep spec template concise in SKILL.md
  • Reference detailed templates in references/ directory
  • Load templates only when needed

IEEE 29148:2018 Quality Assessment Framework

The requirements quality score is calculated from 5 sections, each containing specific criteria with maximum weight of 5 per criterion.

Section 1: Core Quality Attributes (40 points total)

CriterionWeightCheck
Necessary5Does the requirement define an essential capability/characteristic?
Unambiguous5Is there only one possible interpretation?
Complete5Are all necessary capabilities/characteristics/constraints described?
Consistent5Is it free of conflicts with other requirements?
Singular5Does it address only one concern (not compound)?
Feasible5Is it achievable with given constraints (time, budget, technology)?
Traceable5Can it be linked to stakeholder needs and business objectives?
Verifiable5Can fulfillment be measured through testing/inspection?

Section 2: Context & Rationale (20 points total)

CriterionWeightCheck
Implementation-free5Does it avoid dictating HOW to implement (WHAT vs HOW)?
Affordable5Is it within budget/resource constraints?
Bounded5Are scope boundaries clearly defined?
Rationale stated5Is the reason for this requirement explained?

Section 3: Completeness of Description (20 points total)

CriterionWeightCheck
Inputs defined5Are all inputs clearly specified (data, parameters, user actions)?
Outputs defined5Are expected outputs clearly described (return values, UI changes, side effects)?
Constraints identified5Are technical/business/environmental constraints noted?
Edge cases covered5Are error scenarios and boundary conditions addressed?

Section 4: Traceability & Stakeholders (10 points total)

CriterionWeightCheck
Stakeholder mapping5Can the requirement be linked to specific stakeholders/users?
Business value5Does it align with stated business objectives or user needs?

Section 5: Acceptance & Prioritization (10 points total)

CriterionWeightCheck
Prioritization clear5Is priority level specified (must-have vs nice-to-have)?
Success criteria5Are measurable acceptance criteria defined?

Maximum Score: 100 points Threshold: 85 points minimum to proceed

Iterative Validation Process

Round Structure

For each validation round:

  1. Display Current Requirements Version - Show the accumulated requirements in markdown format
  2. Calculate IEEE 29148 Score - Evaluate each criterion and calculate total
  3. Decision Point:

- If score ≥ 85: Proceed to next phase - If score < 85: Use AskUserQuestion to clarify gaps

  1. Enrich & Re-evaluate - Update requirements with answers and re-score

Round 1: Core Requirements Validation

Initial Requirements Gathering:

## Initial Requirements

### Feature Overview
- **Feature name**: {from user}
- **Type**: {new feature/bug fix/enhancement}
- **Vision**: {one sentence}

### Core Requirements
- {Must-have requirements from user}

### Stakeholders
- {Target users, roles}

### Constraints
- {Technical/business constraints}

Evaluate using Section 1 (Core Quality Attributes) + Section 2 (Context & Rationale)

Common Gaps to Address via AskUserQuestion:

  • Unambiguous: "Which specific notification channels?" (email, in-app, SMS?)
  • Complete: "What data needs to be stored for each notification?"
  • Feasible: "Are there rate limits or API constraints?"
  • Implementation-free: "What business outcome, not which library?"

Round 2: Technical Completeness Validation

Extended Requirements:

## Technical Requirements

### Functional Requirements
- {Specific functional behaviors}

### Non-Functional Requirements
- {Performance, security, usability}

### API & Database Changes
- {Endpoints, schemas, migrations}

### Frontend Changes
- {Pages, components, templates}

Evaluate using Section 3 (Completeness of Description)

Common Gaps to Address:

  • Inputs defined: "What parameters does the API accept?"
  • Outputs defined: "What response format/status codes?"
  • Constraints identified: "Are there authentication requirements?"
  • Edge cases covered: "What happens if the external service is down?"

Round 3: Traceability & Acceptance Validation

Final Requirements:

## Implementation Plan

### Stages
- {Logical breakdown with dependencies}

### Acceptance Criteria per Stage
- {Measurable, testable criteria}

### File Changes
- {Concrete paths: src/...}

### Documentation Updates
- {ARCHITECTURE.md, API docs, etc.}

Evaluate using Section 4 (Traceability) + Section 5 (Acceptance)

Common Gaps to Address:

  • Stakeholder mapping: "Which user role triggers this workflow?"
  • Business value: "What problem does this solve for the user?"
  • Success criteria: "How do we verify this works?" (not "it works")
  • Prioritization: "Is this must-have for MVP or can it wait?"

Output Format

After all validation rounds pass, generate the specification file:

# {Feature Name} - Feature Specification

> **Дата создания:** {YYYY-MM-DD}
> **Ветка:** `feature/{slug}`
> **Статус:** В разработке

---

## Видение

{Vision statement}

**Ключевые требования:**
- {Bullet list of must-haves}

---

## План реализации

### Этап 1: {Stage Name} (~{lines} строк)
**Статус:** В разработке

**Цель:** {Purpose}

**Задачи:**
- [ ] {Task 1}
- [ ] {Task 2}

**Файлы:**
- `path/to/file.py` (create/modify)

**Критерий приёмки:**
- {Testable criterion}

---

## История изменений

| Дата | Этап | Коммит | Описание |
| ---- | ---- | ------ | -------- |
| {date} | - | - | ТЗ создано |

Validation Examples

Example 1: Ambiguous Requirement

Initial: "System should be fast"

IEEE 29148 Evaluation:

  • Unambiguous: 0/5 (what is "fast"?)
  • Verifiable: 0/5 (how to measure?)
  • Score: 2/100

AskUserQuestion:

What performance target do you need?
- Response time < 200ms
- Throughput > 100 req/sec
- Page load < 3s

Refined: "API endpoint must respond within 200ms for 95th percentile"

IEEE 29148 Evaluation:

  • Unambiguous: 5/5 (clear metric)
  • Verifiable: 5/5 (can measure)
  • Score: 90/100

Example 2: Incomplete Requirement

Initial: "Users receive notifications"

IEEE 29148 Evaluation:

  • Complete: 1/5 (missing what, when, how)
  • Inputs defined: 0/5
  • Score: 15/100

After AskUserQuestion rounds: "Logged-in users receive in-app notifications when wallet balance changes, with notification storing: user_id, message, timestamp, is_read flag"

IEEE 29148 Evaluation:

  • Complete: 5/5
  • Inputs defined: 5/5
  • Outputs defined: 5/5
  • Verifiable: 5/5
  • Score: 88/100

Required Templates by Round

Round 1 Questions (Core Requirements)

  1. What specific problem does this solve for the user?
  2. Who are the target users/stakeholders? (roles, permissions)
  3. What are the essential capabilities (must-haves)?
  4. What defines "done" for this feature?
  5. Are there technical/business constraints?

Round 2 Questions (Technical Details)

  1. Which modules/services are affected?
  2. What database changes are needed (tables, migrations)?
  3. What API endpoints (new or modified)?
  4. What frontend pages/components change?
  5. What external dependencies or integrations?

Round 3 Questions (Implementation)

  1. Can this be broken into 3-9 logical stages?
  2. What are the dependencies between stages?
  3. What are measurable acceptance criteria for each stage?
  4. Which files will be modified (concrete paths)?

Reference Templates

Common Pitfalls

  1. Accepting vague requirements: "Make it faster" → "What latency target?"
  2. Compound requirements: "User logs in and sees dashboard" → Split into two
  3. Implementation details in requirements: "Use Redis" → "Cache with <100ms latency"
  4. Skipping rounds: Must complete ALL 3 validation rounds with ≥85%
  5. Generic acceptance criteria: "It works" → "API returns 200 with valid JSON schema"

Example Workflow

User: "Создай ТЗ для новой фичи уведомлений"

Claude: [Displays Initial Requirements]
## Initial Requirements
- Feature: Notifications
- Type: New feature
- Vision: Users receive notifications for events
- Must-have: Notifications for events

Claude: [Evaluates - Score: 18/100]
❌ Unambiguous: 0/5 (what events? what channels?)
❌ Complete: 1/5 (missing what, when, how)
❌ Verifiable: 0/5 (no acceptance criteria)

Claude: [Uses AskUserQuestion]
1. What types of events trigger notifications? → "Wallet changes, meeting reminders"
2. Which notification channels? → "Email + in-app"
3. Who receives notifications? → "Logged-in users, based on their preferences"

Claude: [Re-displays Requirements v2]
## Requirements v2
- Event types: wallet_balance_change, meeting_reminder
- Channels: email, in-app
- Recipients: users based on preferences

Claude: [Re-evaluates - Score: 42/100]
✅ Stakeholder mapping: 5/5
✅ Necessary: 5/5
❌ Inputs defined: 2/5 (what data in notification?)
❌ Outputs defined: 2/5 (what does user see?)
❌ Edge cases: 0/5

Claude: [Uses AskUserQuestion - Round 2]
1. What data for each notification type?
2. What if email service is down?
3. How are notifications stored/displayed?

[... continues until score ≥85 ...]

Claude: [Final Score: 87/100] ✅
Proceeds to generate .ai/specs/feature-notifications.md

Final Validation Checklist

Before saving specification, verify:

  1. ✅ All 3 validation rounds passed with ≥85% score
  2. ✅ File follows project spec format
  3. ✅ Each stage has concrete file paths and acceptance criteria
  4. ✅ Russian language used for vision/tasks/criteria
  5. ✅ Technical terms in English (API, database, etc.)
  6. ✅ Requirements are implementation-free (WHAT, not HOW)

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

04

需要参考平台分布和安装热度时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

trae

61.99%
按下载量换算44

Claude Code

27.42%
按下载量换算19

安全审计

暂无安全审计结果可展示。

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills