Cc/skill handbook workflow default

README.md

@@ -51,8 +51,15 @@ node bin/stylekit-style-prompts-skill.js uninstall --target /tmp/stylekit-skill-
 python3 scripts/audit_style_rule_conflicts.py --format text
 python3 scripts/smoke_test.py
 python3 scripts/run_pipeline.py --query "高端科技SaaS财务后台，玻璃质感，强调可读性" --stack nextjs --format json
+python3 scripts/run_pipeline.py --workflow codegen --query "高端科技SaaS财务后台，玻璃质感，强调可读性" --stack nextjs --format json
 ```
 
+说明：
+- 默认是 `--workflow manual`（手册/知识库模式）：输出设计简报 + 手册化建议，不强制走 prompt QA。
+- 若要生成并严格审查 prompt，请显式加 `--workflow codegen`。
+- 在 manual 模式下，会额外输出 `manual_assistant.decision_assistant`，包含：候选风格卡片、给新手的引导问题、以及用户选定风格后的下一步命令模板。
+- 可直接复用对话模板：`references/cc-decision-conversation-template.md`。
+
 ## 3) 回归门禁
 
 ```bash

SKILL.md

@@ -11,29 +11,33 @@ Generate better-looking frontend output by combining StyleKit style identity, ac
 
 ## Quick One-shot Command
 
-Run full flow in one command:
+Run handbook mode in one command (default):
 
 `python scripts/run_pipeline.py --query "<requirement>" --stack nextjs --format json`
 
+Run prompt-generation mode with QA gate:
+
+`python scripts/run_pipeline.py --workflow codegen --query "<requirement>" --stack nextjs --format json`
+
 Force multi-style blend:
 
-`python scripts/run_pipeline.py --query "<requirement>" --stack nextjs --blend-mode on --format json`
+`python scripts/run_pipeline.py --workflow codegen --query "<requirement>" --stack nextjs --blend-mode on --format json`
 
 Run targeted refinement (polish/debug/contrast/layout/component-fill):
 
-`python scripts/run_pipeline.py --query "<requirement>" --stack nextjs --refine-mode debug --format json`
+`python scripts/run_pipeline.py --workflow codegen --query "<requirement>" --stack nextjs --refine-mode debug --format json`
 
 Run with screenshot/Figma reference constraints:
 
-`python scripts/run_pipeline.py --query "<requirement>" --stack nextjs --reference-type screenshot --reference-notes "<what to preserve/fix>" --format json`
+`python scripts/run_pipeline.py --workflow codegen --query "<requirement>" --stack nextjs --reference-type screenshot --reference-notes "<what to preserve/fix>" --format json`
 
 Run with structured reference payload:
 
-`python scripts/run_pipeline.py --query "<requirement>" --stack nextjs --reference-type screenshot --reference-file refs/screen-analysis.json --format json`
+`python scripts/run_pipeline.py --workflow codegen --query "<requirement>" --stack nextjs --reference-type screenshot --reference-file refs/screen-analysis.json --format json`
 
 Run strict schema mode for reference payload:
 
-`python scripts/run_pipeline.py --query "<requirement>" --stack nextjs --reference-type screenshot --reference-file refs/screen-analysis.json --strict-reference-schema --format json`
+`python scripts/run_pipeline.py --workflow codegen --query "<requirement>" --stack nextjs --reference-type screenshot --reference-file refs/screen-analysis.json --strict-reference-schema --format json`
 
 Benchmark current quality:
 
@@ -75,6 +79,17 @@ CI one-command gate:
    `python scripts/generate_brief.py --query "<requirement>" --stack nextjs --mode brief+prompt --reference-file refs/screen-analysis.json --strict-reference-schema`
 9. If quality gate fails, run audit + fix workflow.
 
+## Workflow 1.5: Novice Decision Assist (Recommended)
+
+1. User only provides a high-level goal (example: "I want to build a blog").
+2. Run handbook mode:
+   `python scripts/run_pipeline.py --query "<requirement>" --stack nextjs --format json`
+3. Read `manual_assistant.decision_assistant.recommended_style_options` and explain 3-4 options with trade-offs.
+4. Ask `manual_assistant.decision_assistant.decision_questions` to help user pick direction.
+5. After user selects one option, run codegen mode with forced style:
+   `python scripts/run_pipeline.py --workflow codegen --query "<requirement>" --stack nextjs --style <slug> --blend-mode off --format json`
+6. Follow `references/cc-decision-conversation-template.md` for a turn-by-turn assistant script.
+
 ## Workflow 2: Existing Prompt -> Quality Audit -> Fix Suggestions
 
 1. Audit prompt text:
@@ -149,6 +164,7 @@ If stack is unknown, fallback to framework-agnostic Tailwind semantics.
 - `references/frontend-design-principles.md`: distinctiveness and anti-generic design heuristics.
 - `references/design-system-patterns.md`: token hierarchy and component architecture.
 - `references/accessibility-gate.md`: WCAG + mobile touch baseline for prompt quality.
+- `references/cc-decision-conversation-template.md`: assistant dialogue template for novice user decision flow.
 - `scripts/refresh-style-prompts.sh`: rebuild style dataset from local repo.
 - `scripts/search_stylekit.py`: query -> ranked style candidates.
 - `scripts/generate_brief.py`: query -> design brief + prompts.

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@anxforever/stylekit-style-prompts-skill",
-  "version": "0.1.0",
+  "version": "0.1.1",
   "description": "Standalone StyleKit style-prompts skill installer for Codex and Claude.",
   "license": "MIT",
   "type": "commonjs",

references/cc-decision-conversation-template.md

@@ -0,0 +1,52 @@
+# CC Decision Conversation Template
+
+Use this when the user only has a high-level idea (for example: "I want a blog").
+
+## Goal
+
+Help non-frontend users make design decisions first, then generate implementation.
+
+## Phase 1: Clarify Product Context
+
+- Assistant prompt:
+  - "I will help narrow direction first. Who is the audience, and is the primary goal reading, conversion, or brand expression?"
+- Expected output:
+  - product type
+  - target audience
+  - top 1-2 business goals
+
+## Phase 2: Present Style Choices
+
+- Input source:
+  - `manual_assistant.decision_assistant.recommended_style_options`
+- Assistant prompt:
+  - "Here are 3-4 style options with complexity and risk. Pick A/B/C/D, and I will generate with that style."
+- Explain each option in plain language:
+  - visual tone
+  - implementation complexity
+  - readability/risk trade-off
+
+## Phase 3: Ask Guiding Questions
+
+- Input source:
+  - `manual_assistant.decision_assistant.decision_questions`
+- Ask 2-3 questions max:
+  - visual intensity
+  - content density preference
+  - motion level
+
+## Phase 4: Lock Style and Generate
+
+- After user picks a style option, run:
+  - `python scripts/run_pipeline.py --workflow codegen --query "<requirement>" --stack <stack> --style <slug> --blend-mode off --format json`
+- Then generate implementation with:
+  - homepage
+  - detail page (for blog: post page)
+  - list page (for blog: article list page)
+
+## Output Rules for Assistant
+
+- Never force a style before user selects.
+- Explain style differences in non-technical language first.
+- Preserve user language (Chinese in -> Chinese out, English in -> English out).
+- Keep decision loop short: propose options -> ask questions -> confirm -> generate.

scripts/benchmark_pipeline.py

@@ -123,6 +123,8 @@ def run_case(py: str, query: str, stack: str, blend_mode: str, bucket: str, refi
     cmd = [
         py,
         str(PIPELINE_SCRIPT),
+        "--workflow",
+        "codegen",
         "--query",
         query,
         "--stack",