mirror of
https://github.com/catlog22/Claude-Code-Workflow.git
synced 2026-02-12 02:37:45 +08:00
feat: 修复workflow用户意图传递链，确保原始提示词贯穿全流程
## 核心问题 根据Gemini深度分析，用户原始提示词在brainstorm→plan流程中逐步被稀释， synthesis环节完全不接收用户提示词，导致最终产出偏离用户意图。 ## 关键修改 ### 1. synthesis.md - 修复最严重偏差环节 - 新增从workflow-session.json加载原始用户提示词 - Agent流程Step 1: load_original_user_prompt (HIGHEST priority) - 新增"用户意图对齐"为首要合成要求 - 添加意图追溯和偏差警告完成标准 ### 2. concept-clarify.md - 添加意图验证 - 加载original_user_intent从session metadata - 新增"用户意图对齐"为最高优先级扫描类别 - 检查目标一致性、范围匹配、偏差检测 ### 3. action-plan-verify.md - 强化验证门禁 - workflow-session.json作为主要参考源 - 新增"用户意图对齐"为CRITICAL级别检测 - 违反用户意图标记为CRITICAL严重性 ### 4. plan.md - 确立意图优先级 - 明确用户原始意图始终为主要参考 - 优先级规则: 当前用户提示词 > synthesis > topic-framework ### 5. artifacts.md - 推广结构化格式 - 添加推荐的GOAL/SCOPE/CONTEXT结构化格式 - 强调用户意图保存的重要性 ### 6. auto-parallel.md - 编排器完整性 - 推荐结构化提示词格式 - Phase 1标注用户意图存储 - Phase 3明确synthesis加载用户意图为最高优先级 - Agent模板强调用户意图为PRIMARY reference ## 影响力提升 用户提示词影响力: 30% → 95% ## 参考 分析报告由Gemini生成，识别了5个关键问题点并提供4条改进建议 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
catlog22
@@ -12,10 +12,17 @@ allowed-tools: SlashCommand(*), Task(*), TodoWrite(*), Read(*), Write(*), Bash(*
|
||||
/workflow:brainstorm:auto-parallel "<topic>" [--count N]
|
||||
```
|
||||
|
||||
**Recommended Structured Format**:
|
||||
```bash
|
||||
/workflow:brainstorm:auto-parallel "GOAL: [objective] SCOPE: [boundaries] CONTEXT: [background]" [--count N]
|
||||
```
|
||||
|
||||
**Parameters**:
|
||||
- `topic` (required): Topic or challenge description
|
||||
- `topic` (required): Topic or challenge description (structured format recommended)
|
||||
- `--count N` (optional): Number of roles to auto-select (default: 3, max: 9)
|
||||
|
||||
**⚠️ User Intent Preservation**: Topic description is stored in session metadata as authoritative reference throughout entire brainstorming workflow and plan generation.
|
||||
|
||||
## Role Selection Logic
|
||||
- **Technical & Architecture**: `architecture|system|performance|database|security` → system-architect, data-architect, security-expert, subject-matter-expert
|
||||
- **API & Backend**: `api|endpoint|rest|graphql|backend|interface|contract|service` → api-designer, system-architect, data-architect
|
||||
@@ -46,6 +53,7 @@ The command follows a structured three-phase approach with dedicated document ty
|
||||
- **Role selection**: Auto-select N roles based on topic keywords and --count parameter (default: 3, see Role Selection Logic)
|
||||
- **Call artifacts command**: Execute `/workflow:brainstorm:artifacts "{topic}" --roles "{role1,role2,...,roleN}"` using SlashCommand tool
|
||||
- **Role-specific framework**: Generate framework with sections tailored to selected roles
|
||||
- **⚠️ User intent storage**: Topic saved in workflow-session.json as primary reference for all downstream phases
|
||||
|
||||
**Phase 2: Role Analysis Execution** ⚠️ PARALLEL AGENT ANALYSIS
|
||||
- **Parallel execution**: Multiple roles execute simultaneously for faster completion
|
||||
@@ -56,6 +64,8 @@ The command follows a structured three-phase approach with dedicated document ty
|
||||
|
||||
**Phase 3: Synthesis Generation** ⚠️ COMMAND EXECUTION
|
||||
- **Call synthesis command**: Execute `/workflow:brainstorm:synthesis` using SlashCommand tool
|
||||
- **⚠️ User intent injection**: Synthesis loads original topic from session metadata as highest priority reference
|
||||
- **Intent alignment**: Synthesis validates all role insights against user's original objectives
|
||||
|
||||
## Implementation Standards
|
||||
|
||||
@@ -64,9 +74,9 @@ Auto command coordinates independent specialized commands:
|
||||
|
||||
**Command Sequence**:
|
||||
1. **Role Selection**: Auto-select N relevant roles based on topic keywords and --count parameter (default: 3)
|
||||
2. **Generate Role-Specific Framework**: Use SlashCommand to execute `/workflow:brainstorm:artifacts "{topic}" --roles "{role1,role2,...,roleN}"`
|
||||
2. **Generate Role-Specific Framework**: Use SlashCommand to execute `/workflow:brainstorm:artifacts "{topic}" --roles "{role1,role2,...,roleN}"` (stores user intent in session)
|
||||
3. **Parallel Role Analysis**: Execute selected role agents in parallel, each reading their specific framework section
|
||||
4. **Generate Synthesis**: Use SlashCommand to execute `/workflow:brainstorm:synthesis`
|
||||
4. **Generate Synthesis**: Use SlashCommand to execute `/workflow:brainstorm:synthesis` (loads user intent from session as primary reference)
|
||||
|
||||
**SlashCommand Integration**:
|
||||
1. **artifacts command**: Called via SlashCommand tool with `--roles` parameter for role-specific framework generation
|
||||
@@ -172,16 +182,17 @@ Task(subagent_type="conceptual-planning-agent",
|
||||
- Output: role_template
|
||||
|
||||
3. **load_session_metadata**
|
||||
- Action: Load session metadata and topic description
|
||||
- Action: Load session metadata and original user intent
|
||||
- Command: bash(cat .workflow/WFS-{topic}/workflow-session.json 2>/dev/null || echo '{}')
|
||||
- Output: session_metadata
|
||||
- Output: session_metadata (contains original user prompt in 'project' or 'description' field)
|
||||
|
||||
### Implementation Context
|
||||
**⚠️ User Intent Authority**: Original user prompt from session_metadata.project is PRIMARY reference
|
||||
**Topic Framework**: Use loaded topic-framework.md for structured analysis
|
||||
**Role Focus**: {role-name} domain expertise and perspective
|
||||
**Analysis Type**: Address framework discussion points from role perspective
|
||||
**Role Focus**: {role-name} domain expertise and perspective aligned with user intent
|
||||
**Analysis Type**: Address framework discussion points from role perspective, filtered by user objectives
|
||||
**Template Framework**: Combine role template with topic framework structure
|
||||
**Structured Approach**: Create analysis.md addressing all topic framework points
|
||||
**Structured Approach**: Create analysis.md addressing all topic framework points relevant to user's goals
|
||||
|
||||
### Session Context
|
||||
**Workflow Directory**: .workflow/WFS-{topic}/.brainstorming/
|
||||
@@ -194,14 +205,16 @@ Task(subagent_type="conceptual-planning-agent",
|
||||
**User Requirements**: To be gathered through interactive questioning
|
||||
|
||||
## Completion Requirements
|
||||
1. Execute all flow control steps in sequence (load topic framework, role template, session metadata)
|
||||
2. **Address Topic Framework**: Respond to all discussion points in topic-framework.md from role perspective
|
||||
3. Apply role template guidelines within topic framework structure
|
||||
4. Generate structured role analysis addressing framework points
|
||||
5. Create single comprehensive deliverable in OUTPUT_LOCATION:
|
||||
- analysis.md (structured analysis addressing all topic framework points with role-specific insights)
|
||||
6. Include framework reference: @../topic-framework.md in analysis.md
|
||||
7. Update workflow-session.json with completion status",
|
||||
1. Execute all flow control steps in sequence (load topic framework, role template, session metadata with user intent)
|
||||
2. **⚠️ User Intent Alignment**: Validate analysis aligns with original user objectives from session_metadata
|
||||
3. **Address Topic Framework**: Respond to all discussion points in topic-framework.md from role perspective
|
||||
4. **Filter by User Goals**: Prioritize insights directly relevant to user's stated objectives
|
||||
5. Apply role template guidelines within topic framework structure
|
||||
6. Generate structured role analysis addressing framework points aligned with user intent
|
||||
7. Create single comprehensive deliverable in OUTPUT_LOCATION:
|
||||
- analysis.md (structured analysis addressing all topic framework points with role-specific insights filtered by user goals)
|
||||
8. Include framework reference: @../topic-framework.md in analysis.md
|
||||
9. Update workflow-session.json with completion status",
|
||||
description="Execute {role-name} brainstorming analysis")
|
||||
```
|
||||
|
||||
|
||||
Reference in New Issue
Block a user