skillkit
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseSection 1: Intent Detection & Routing
第1部分:意图检测与路由
Detect user intent, route to appropriate workflow.
| Intent | Keywords | Route To |
|---|---|---|
| Full skill creation | "create skill", "build skill", "new skill" | Section 2 |
| Subagent creation | "create subagent", "build subagent", "new subagent" | Section 6 |
| Validation | "validate", "check quality" | Section 3 |
| Decision | "Skills vs Subagents", "decide", "which to use" | Section 4 |
| Migration | "convert", "migrate doc" | Section 5 |
| Single tool | "validate only", "estimate tokens", "scan" | Section 7 |
PROCEED to corresponding section after intent detection.
Stop Condition (Mandatory):
- If multiple routes match or intent is ambiguous: stop, ask user to choose one route.
- Do not proceed until user confirms the route.
Workflow Value: Research-driven approach validates design before building.
Sequential steps with checkpoints produce 9.0/10+ quality vs ad-hoc creation.
检测用户意图,路由至对应的工作流。
| 意图 | 关键词 | 路由至 |
|---|---|---|
| 完整Skill创建 | "create skill", "build skill", "new skill" | 第2部分 |
| Subagent创建 | "create subagent", "build subagent", "new subagent" | 第6部分 |
| 验证 | "validate", "check quality" | 第3部分 |
| 决策 | "Skills vs Subagents", "decide", "which to use" | 第4部分 |
| 迁移 | "convert", "migrate doc" | 第5部分 |
| 单一工具 | "validate only", "estimate tokens", "scan" | 第7部分 |
检测到意图后,进入对应的部分。
终止条件(强制要求):
- 若多个路由匹配或意图不明确:终止流程,请求用户选择一个路由。
- 获得用户确认前,不得继续执行。
工作流价值: 基于研究的方法可在构建前验证设计。
带检查点的分步流程相比临时创建,可产出9.0/10以上的质量。
Section 2: Creation Workflows (Dual Mode)
第2部分:创建工作流(双模式)
Prerequisites: Skill description provided, workspace available.
前置条件: 已提供Skill描述,工作区可用。
Mode Selection (Required at Start)
模式选择(开始时必须进行)
Detect or prompt for workflow mode before running the creation flow.
Stop Condition (Mandatory):
- If mode is not explicitly provided: stop and ask "Do you want fast or full mode?"
- Do not continue until user confirms the mode.
| Mode | Steps | Validation | Quality Target | Time |
|---|---|---|---|---|
| fast | 10 | Structural only | >=9.0/10 | <10 min |
| full | 14 | Structural + Behavioral | >=9.0/10 and behavioral >=7.0 | <20 min |
No implicit default mode is allowed when mode is not explicitly known.
在运行创建流程前,检测或提示用户选择工作流模式。
终止条件(强制要求):
- 若未明确指定模式:终止流程并询问“您想要使用快速模式还是完整模式?”
- 获得用户确认前,不得继续执行。
| 模式 | 步骤数 | 验证内容 | 质量目标 | 耗时 |
|---|---|---|---|---|
| 快速模式 | 10 | 仅结构验证 | >=9.0/10 | <10分钟 |
| 完整模式 | 14 | 结构+行为验证 | >=9.0/10且行为评分>=7.0 | <20分钟 |
当模式未明确时,不允许使用默认模式。
Workflow A: Fast Mode (10 Steps)
工作流A:快速模式(10步)
Use when contains or marker does not exist.
.skillkit-modefast→ READ IN FULL before starting.
Create a task for each step listed in that file, then follow them in order.
The outline below is a summary only — the reference file is authoritative.
references/section-2-fast-creation-workflow.mdPhase 1: Decision & Research
- Step 0: Decide approach ()
decision_helper.py - Step 1: Research and proposals
- Step 2: User validation
- Stop Condition: Stop and request user approval before continuing to Step 3.
Phase 2: Creation
- Step 3: Initialize skill ()
init.py skill <name> --mode fast - Step 4: Create content
Phase 3: Structural Validation
- Step 5: Validate skill () — runs structure + security + tokens in one call
validate_skill.py
Phase 4: Packaging
- Step 6: Progressive disclosure check
- Step 7: Generate tests ()
test_generator.py - Step 8: Quality assessment ()
quality_scorer.py - Step 9: Package ()
package_skill.py
当文件包含或不存在该文件时使用。
.skillkit-modefast→ 开始前,请完整阅读。
为该文件中列出的每个步骤创建任务,然后按顺序执行。
以下大纲仅为摘要——参考文件为权威依据。
references/section-2-fast-creation-workflow.md阶段1:决策与研究
- 步骤0:确定方法()
decision_helper.py - 步骤1:研究与提案
- 步骤2:用户验证
- 终止条件:进入步骤3前,终止流程并请求用户批准。
阶段2:创建
- 步骤3:初始化Skill()
init.py skill <name> --mode fast - 步骤4:创建内容
阶段3:结构验证
- 步骤5:验证Skill()——一次性运行结构、安全与Token验证
validate_skill.py
阶段4:打包
- 步骤6:渐进式披露检查
- 步骤7:生成测试用例()
test_generator.py - 步骤8:质量评估()
quality_scorer.py - 步骤9:打包()
package_skill.py
Workflow B: Full Mode (14 Steps)
工作流B:完整模式(14步)
Use when contains .
.skillkit-modefull→ READ IN FULL before starting.
Create a task for each step listed in that file, then follow them in order.
The outline below is a summary only — the reference file is authoritative.
references/section-2-full-creation-workflow.mdPhase 1: Decision and Research
- Step 0: Decide approach ()
decision_helper.py - Step 1: Research and proposals
- Step 2: User validation
- Stop Condition: Stop and request user approval before continuing to Step 3.
Phase 2: Behavioral Baseline (extra vs fast)
- Step 3 (RED): Run pressure scenarios without skill
→ Load → section "Full Mode Behavioral Testing Protocol" (mandatory)
references/section-2-full-creation-workflow.md - Step 4: Document baseline failures
Phase 3: Creation
- Step 5: Initialize skill ()
init.py skill <name> --mode full - Step 6: Create content addressing baseline failures
Phase 4: Behavioral Verification (extra vs fast)
- Step 7 (GREEN): Run scenarios with skill
→ Load → section "Full Mode Behavioral Testing Protocol" (mandatory)
references/section-2-full-creation-workflow.md - Step 8: Fix gaps
Phase 5: Structural Validation
- Step 9: Validate skill () — runs structure + security + tokens in one call
validate_skill.py
Phase 6: Refinement (extra vs fast)
- Step 10 (REFACTOR): Combined pressure tests
→ Load → section "Full Mode Behavioral Testing Protocol" (mandatory)
references/section-2-full-creation-workflow.md - Step 11: Close loopholes
Phase 7: Packaging
- Step 12: Quality assessment () — behavioral score derived from Steps 3/7/10 subagent results, not from
quality_scorer.py --format jsonflag--behavioral - Step 13: Package ()
package_skill.py
当文件包含时使用。
.skillkit-modefull→ 开始前,请完整阅读。
为该文件中列出的每个步骤创建任务,然后按顺序执行。
以下大纲仅为摘要——参考文件为权威依据。
references/section-2-full-creation-workflow.md阶段1:决策与研究
- 步骤0:确定方法()
decision_helper.py - 步骤1:研究与提案
- 步骤2:用户验证
- 终止条件:进入步骤3前,终止流程并请求用户批准。
阶段2:行为基准(相比快速模式新增)
- 步骤3(红色阶段):在无Skill的情况下运行压力场景
→ 加载→ 章节“完整模式行为测试协议”(强制要求)
references/section-2-full-creation-workflow.md - 步骤4:记录基准测试失败情况
阶段3:创建
- 步骤5:初始化Skill()
init.py skill <name> --mode full - 步骤6:创建解决基准测试失败问题的内容
阶段4:行为验证(相比快速模式新增)
- 步骤7(绿色阶段):在有Skill的情况下运行场景
→ 加载→ 章节“完整模式行为测试协议”(强制要求)
references/section-2-full-creation-workflow.md - 步骤8:修复漏洞
阶段5:结构验证
- 步骤9:验证Skill()——一次性运行结构、安全与Token验证
validate_skill.py
阶段6:优化(相比快速模式新增)
- 步骤10(重构阶段):组合压力测试
→ 加载→ 章节“完整模式行为测试协议”(强制要求)
references/section-2-full-creation-workflow.md - 步骤11:填补漏洞
阶段7:打包
- 步骤12:质量评估()——行为评分来自步骤3/7/10的Subagent结果,而非
quality_scorer.py --format json参数--behavioral - 步骤13:打包()
package_skill.py
Mode Detection
模式检测
Priority order:
- Explicit flag: or
--mode fast--mode full - Skill marker: file content
.skillkit-mode - If unknown: stop and ask user to choose or
fastfull
优先级顺序:
- 显式参数:或
--mode fast--mode full - Skill标记:文件内容
.skillkit-mode - 若未知:终止流程并询问用户选择或
fastfull
Section 3: Validation Workflow (Overview)
第3部分:验证工作流(概述)
Use when: Validating existing skill
Steps: Execute validation subset (Steps 3-6)
- Validate skill — structure + security + tokens (, no flags needed)
validate_skill.py - Progressive disclosure check
- Test generation (optional)
- Quality assessment (quality_scorer.py)
Note: and flags are available for Section 7 individual tool use, not for workflow validation steps.
--security-only--tokens-onlyFor detailed workflow: See references/section-3-validation-workflow-existing-skill.md
适用场景: 验证现有Skill
步骤: 执行验证子集(步骤3-6)
- 验证Skill —— 结构+安全+Token(,无需参数)
validate_skill.py - 渐进式披露检查
- 生成测试用例(可选)
- 质量评估()
quality_scorer.py
注意: 和参数仅适用于第7部分的单一工具使用,不适用于工作流验证步骤。
--security-only--tokens-only详细工作流: 参考references/section-3-validation-workflow-existing-skill.md
Section 4: Decision Workflow (Overview)
第4部分:决策工作流(概述)
Use when: Uncertain if Skills is right approach
CRITICAL: Agent MUST create a temp JSON file first. The script does NOT accept inline JSON strings - it requires a file path to a JSON file.
decision_helper.pyStep-by-step invocation: See
references/section-4-decision-workflow-skills-vs-subagents.mdAccuracy: Highest (90-95% confidence).
Process:
- Run with json file.
decision_helper.py - Answer interactive questions
- Receive recommendation with confidence score
- Proceed if Skills recommended (confidence >=75%)
- If confidence <75% or recommendation is uncertain, stop and ask user whether to continue, switch route, or refine inputs.
For detailed workflow: See references/section-4-decision-workflow-skills-vs-subagents.md
适用场景: 不确定是否应使用Skill的情况
关键要求:Agent必须先创建临时JSON文件。 脚本不接受内联JSON字符串——它需要指向JSON文件的路径。
decision_helper.py分步调用说明: 参考
references/section-4-decision-workflow-skills-vs-subagents.md准确率: 最高(90-95%置信度)。
流程:
- 使用JSON文件运行。
decision_helper.py - 回答交互式问题
- 接收带置信度评分的推荐结果
- 若推荐使用Skill且置信度>=75%,则继续执行
- 若置信度<75%或推荐结果不确定,终止流程并询问用户是否继续、切换路由或优化输入。
详细工作流: 参考references/section-4-decision-workflow-skills-vs-subagents.md
Section 6: Subagent Creation Workflow (Overview)
第6部分:Subagent创建工作流(概述)
Use when: Creating new subagent (user explicitly asks or decision workflow recommends)
Prerequisites: Role definition clear, workspace available
Quality Target: Clear role, comprehensive workflow, testable examples
Time: <15 min with template
适用场景: 创建新Subagent(用户明确要求或决策工作流推荐)
前置条件: 角色定义清晰,工作区可用
质量目标: 角色明确、工作流全面、示例可测试
耗时: 使用模板时<15分钟
8-Step Process:
8步流程:
STEP 0: Requirements & Role Definition
- Answer: Primary role? Trigger conditions? Tool requirements?
- Choose subagent_type from predefined list
STEP 1: Initialize Subagent File
- Tool:
python scripts/init.py subagent subagent-name --path ~/.claude/agents - Creates: with template
~/.claude/agents/subagent-name.md - Important: Subagents are individual files (not directories)
.md - Stop Condition: If target file already exists, stop and ask whether to overwrite, rename, or cancel.
STEP 2: Define Configuration
- Edit YAML frontmatter (name, description, type, tools, skills)
- Configure tool permissions (minimal but sufficient)
STEP 3: Define Role and Workflow
- Role definition section
- Trigger conditions (when to invoke)
- Multi-phase workflow
STEP 4: Define Response Format
- Output structure template
- Tone and style guidelines
- Error handling
STEP 5: Add Examples
- At least 1 complete example
- Input/Process/Output format
STEP 6: Validation
- YAML validity check
- Structure verification
- Completeness review
STEP 7: Testing
- Test invocation with Task tool
- Iterate based on results
STEP 8: Documentation & Deployment
- Create README.md
- Register in system
- Stop Condition: Ask for explicit user confirmation before register/deploy actions.
For detailed workflow: See references/section-6-subagent-creation-workflow.md
步骤0:需求与角色定义
- 明确:主要角色?触发条件?工具需求?
- 从预定义列表中选择subagent_type
步骤1:初始化Subagent文件
- 工具:
python scripts/init.py subagent subagent-name --path ~/.claude/agents - 创建:文件(带模板)
~/.claude/agents/subagent-name.md - 重要提示: Subagent是独立的文件(而非目录)
.md - 终止条件:若目标文件已存在,终止流程并询问用户是否覆盖、重命名或取消。
步骤2:定义配置
- 编辑YAML前置元数据(名称、描述、类型、工具、Skills)
- 配置工具权限(最小但足够)
步骤3:定义角色与工作流
- 角色定义章节
- 触发条件(何时调用)
- 多阶段工作流
步骤4:定义响应格式
- 输出结构模板
- 语气与风格指南
- 错误处理
步骤5:添加示例
- 至少1个完整示例
- 输入/流程/输出格式
步骤6:验证
- YAML有效性检查
- 结构验证
- 完整性审查
步骤7:测试
- 使用Task工具调用测试
- 根据结果迭代优化
步骤8:文档与部署
- 创建README.md
- 在系统中注册
- 终止条件:注册/部署前,请求用户明确确认。
详细工作流: 参考references/section-6-subagent-creation-workflow.md
Section 5: Migration Workflow (Overview)
第5部分:迁移工作流(概述)
Use when: Converting document to skill
Process:
- Decision check (Step 0)
- Migration analysis (migration_helper.py)
- Structure creation
- Execute validation steps (5-8)
- Package (Step 9)
Stop Condition (Mandatory):
- Before structure creation or any write/overwrite operation: ask user confirmation.
- Do not modify files until user confirms.
For detailed workflow: See references/section-5-migration-workflow-doc-to-skill.md
适用场景: 将文档转换为Skill
流程:
- 决策检查(步骤0)
- 迁移分析()
migration_helper.py - 结构创建
- 执行验证步骤(5-8)
- 打包(步骤9)
终止条件(强制要求):
- 在创建结构或执行任何写入/覆盖操作前:请求用户确认。
- 获得用户确认前,不得修改文件。
详细工作流: 参考references/section-5-migration-workflow-doc-to-skill.md
Section 7: Individual Tool Usage
第7部分:单一工具使用
Use when: User needs single tool, not full workflow
Entry Point: User asks for specific tool like "estimate tokens" or "security scan"
适用场景: 用户仅需使用单个工具,而非完整工作流
入口: 用户请求特定工具,如“estimate tokens”或“security scan”
Available Tools
可用工具
Validation Tool:
bash
python scripts/validate_skill.py skill-name/ --format jsonGuide:
knowledge/tools/14-validation-tools-guide.mdToken Estimator:
bash
python scripts/validate_skill.py skill-name/ --tokens-only --format jsonGuide:
knowledge/tools/15-cost-tools-guide.mdSecurity Scanner:
bash
python scripts/validate_skill.py skill-name/ --security-only --format jsonGuide:
knowledge/tools/16-security-tools-guide.mdPattern Detector:
bash
undefined验证工具:
bash
python scripts/validate_skill.py skill-name/ --format json指南:
knowledge/tools/14-validation-tools-guide.mdToken估算工具:
bash
python scripts/validate_skill.py skill-name/ --tokens-only --format json指南:
knowledge/tools/15-cost-tools-guide.md安全扫描工具:
bash
python scripts/validate_skill.py skill-name/ --security-only --format json指南:
knowledge/tools/16-security-tools-guide.md模式检测工具:
bash
undefinedAnalysis mode with JSON output
分析模式(JSON输出)
python scripts/pattern_detector.py "convert PDF to Word" --format json
python scripts/pattern_detector.py "convert PDF to Word" --format json
List all patterns
列出所有模式
python scripts/pattern_detector.py --list --format json
python scripts/pattern_detector.py --list --format json
Interactive mode (text only)
交互模式(仅文本)
python scripts/pattern_detector.py --interactive
Guide: `knowledge/tools/17-pattern-tools-guide.md`
**Decision Helper:**
```bashpython scripts/pattern_detector.py --interactive
指南:`knowledge/tools/17-pattern-tools-guide.md`
**决策辅助工具:**
```bashAnalyze use case (JSON output - agent-layer default)
分析用例(JSON输出 - Agent层默认)
python scripts/decision_helper.py --analyze "code review with validation"
python scripts/decision_helper.py --analyze "code review with validation"
Show decision criteria (JSON output)
显示决策标准(JSON输出)
python scripts/decision_helper.py --show-criteria --format json
python scripts/decision_helper.py --show-criteria --format json
Text mode for human reading (debugging)
人类可读文本模式(调试用)
python scripts/decision_helper.py --analyze "description" --format text
Guide: `knowledge/tools/18-decision-helper-guide.md`
**Test Generator (v1.2: Parameter update):**
```bash
python scripts/test_generator.py skill-name/ --test-format pytest --format json- : Test framework (pytest/unittest/plain, default: pytest)
--test-format - : Output style (text/json, default: text)
--format - Backward compatible: Old parameter still works (deprecated)
--output
Guide:
knowledge/tools/19-test-generator-guide.mdSplit Skill:
bash
python scripts/split_skill.py skill-name/ --format jsonGuide:
knowledge/tools/20-split-skill-guide.mdQuality Scorer:
bash
python scripts/quality_scorer.py skill-name/ --format jsonGuide:
knowledge/tools/21-quality-scorer-guide.mdMigration Helper:
bash
python scripts/migration_helper.py doc.md --format jsonGuide:
knowledge/tools/22-migration-helper-guide.mdSubagent Initializer (NEW):
bash
python scripts/init.py subagent subagent-name --path /path/to/subagentsGuide:
references/section-6-subagent-creation-workflow.mdpython scripts/decision_helper.py --analyze "description" --format text
指南:`knowledge/tools/18-decision-helper-guide.md`
**测试生成工具(v1.2:参数更新):**
```bash
python scripts/test_generator.py skill-name/ --test-format pytest --format json- :测试框架(pytest/unittest/plain,默认:pytest)
--test-format - :输出格式(text/json,默认:text)
--format - 向后兼容:旧参数仍可使用(已废弃)
--output
指南:
knowledge/tools/19-test-generator-guide.mdSkill拆分工具:
bash
python scripts/split_skill.py skill-name/ --format json指南:
knowledge/tools/20-split-skill-guide.md质量评分工具:
bash
python scripts/quality_scorer.py skill-name/ --format json指南:
knowledge/tools/21-quality-scorer-guide.md迁移辅助工具:
bash
python scripts/migration_helper.py doc.md --format json指南:
knowledge/tools/22-migration-helper-guide.mdSubagent初始化工具(新增):
bash
python scripts/init.py subagent subagent-name --path /path/to/subagents指南:
references/section-6-subagent-creation-workflow.mdTool Output Standardization (v1.0.1+)
工具输出标准化(v1.0.1+)
All 9 tools support . Text mode still available via (backward compatible). defaults to JSON for automation.
--format json--format textdecision_helperJSON Output Structure:
json
{
"status": "success" | "error",
"tool": "tool_name",
"timestamp": "ISO-8601",
"data": { /* tool-specific results */ }
}所有9个工具均支持。文本模式仍可通过使用(向后兼容)。默认使用JSON格式以支持自动化。
--format json--format textdecision_helperJSON输出结构:
json
{
"status": "success" | "error",
"tool": "tool_name",
"timestamp": "ISO-8601",
"data": { /* 工具特定结果 */ }
}Quality Assurance Enhancements (v1.2+)
质量保证增强(v1.2+)
File & Reference Validation:
- now comprehensively checks file references (markdown links, code refs, path patterns)
validate_skill.py - validates references before packaging, detects orphaned files
package_skill.py - Prevents broken references and incomplete files in deployed skills
Content Budget Enforcement (v1.2+):
- Hard limits on file size: P0 ≤150 lines, P1 ≤100 lines, P2 ≤60 lines
- Real-time token counting with progress indicators
- Prevents file bloat that previously caused 4-9x target overruns
Execution Planning (v1.2+):
- P0/P1/P2 prioritization prevents over-scoping
- Token budget allocated per file to maintain efficiency
- Research phase respects Verbalized Sampling probability thresholds (p>0.10)
Quality Scorer Context:
- Scores calibrated for general skill quality heuristics
- Target: 70%+ is good, 80%+ is excellent
- Style scoring may not fit all skill types (educational vs technical)
- Use as guidance, supplement with manual review for edge cases
文件与引用验证:
- 现在可全面检查文件引用(Markdown链接、代码引用、路径模式)
validate_skill.py - 在打包前验证引用,检测孤立文件
package_skill.py - 防止已部署Skill中出现无效引用和不完整文件
内容预算管控(v1.2+):
- 文件大小硬限制:P0 ≤150行,P1 ≤100行,P2 ≤60行
- 带进度指示器的实时Token计数
- 防止之前导致4-9倍目标超限的文件膨胀问题
执行规划(v1.2+):
- P0/P1/P2优先级划分可防止范围过度扩大
- 为每个文件分配Token预算以保持效率
- 研究阶段遵循Verbalized Sampling概率阈值(p>0.10)
质量评分工具说明:
- 评分基于通用Skill质量启发式算法校准
- 目标:70%+为良好,80%+为优秀
- 风格评分可能不适用于所有Skill类型(教育类vs技术类)
- 用作指导,针对边缘情况需补充人工审查
Section 8: Mode Selection Guide
第8部分:模式选择指南
| Skill Type | Recommended Mode | Why |
|---|---|---|
| TDD or discipline skill | full | must resist rationalization under pressure |
| Code pattern skill | fast | structural checks are usually sufficient |
| API reference skill | fast | primarily retrieval accuracy |
| Workflow orchestration skill | full | complex flow benefits from pressure checks |
| Debugging technique skill | fast | concise technique with clear method |
Full mode adds behavioral testing (pressure scenarios). Use it when discipline enforcement is core to the skill's purpose.
| Skill类型 | 推荐模式 | 原因 |
|---|---|---|
| TDD或纪律类Skill | 完整模式 | 必须能在压力下抵制合理化行为 |
| 代码模式类Skill | 快速模式 | 结构检查通常已足够 |
| API参考类Skill | 快速模式 | 主要关注检索准确性 |
| 工作流编排类Skill | 完整模式 | 复杂流程可从压力测试中获益 |
| 调试技巧类Skill | 快速模式 | 技巧简洁、方法明确 |
完整模式新增行为测试(压力场景)。当纪律执行为Skill核心目标时使用。
Section 9: Knowledge Reference Map (Overview)
第9部分:知识参考地图(概述)
Strategic context loaded on-demand.
战略上下文按需加载。
Foundation Concepts (Files 01-08):
基础概念(文件01-08):
- Why Skills exist vs alternatives
- Skills vs Subagents decision framework
- Token economics and efficiency
- Platform constraints and security
- When NOT to use Skills
- Skill存在的原因与替代方案对比
- Skill vs Subagent决策框架
- Token经济与效率
- 平台约束与安全
- 不适合使用Skill的场景
Application Knowledge (Files 09-13):
应用知识(文件09-13):
- Real-world case studies (Rakuten, Box, Notion)
- Technical architecture patterns
- Adoption and testing strategies
- Competitive landscape analysis
- 真实案例研究(乐天、Box、Notion)
- 技术架构模式
- 采用与测试策略
- 竞争格局分析
Tool Guides (Files 14-22):
工具指南(文件14-22):
- One guide per automation script
- Usage patterns and parameters
- JSON output formats
- Integration examples
For complete reference map: See references/section-7-knowledge-reference-map.md
- 每个自动化脚本对应一份指南
- 使用模式与参数
- JSON输出格式
- 集成示例
完整参考地图: 参考references/section-7-knowledge-reference-map.md
Workflow Compliance
工作流合规性
Follow workflows sequentially. Sequential steps with gates produce 9.0/10+ quality. Deviations are allowed with user justification.
Flexible entry points:
- Single tool (Section 7): skip full workflow
- Validation only (Section 3): run validation subset
- Subagent (Section 6): streamlined 8-step workflow
按顺序执行工作流。带检查点的分步流程可产出9.0/10以上的质量。经用户说明后可允许偏离。
灵活入口点:
- 单一工具(第7部分):跳过完整工作流
- 仅验证(第3部分):运行验证子集
- Subagent(第6部分):简化的8步工作流
Additional Resources
附加资源
Load reference files on-demand from when detailed implementation guidance is needed.
references/当需要详细实现指导时,从目录按需加载参考文件。
references/