cc-viewer 1.6.232 → 1.6.233

package/concepts/uk/CustomUltraplanExpert.md

@@ -0,0 +1,165 @@
+# Custom UltraPlan Expert — Посібник зі створення
+
+## Що роблять два поля введення
+
+- **Ім'я експерта**: підпис, що відображається на кнопці ролі у рядку варіантів UltraPlan (макс. 30 символів). Це просто відображуване ім'я, і воно **ніколи** не надсилається до Claude Code.
+- **Тіло промпту**: ваша інструкція ролі. Під час надсилання cc-viewer **автоматично** обгортає її в теги `<system-reminder>...</system-reminder>` із заголовком області `[SCOPED INSTRUCTION]`. Тому **пишіть лише тіло** — не додавайте теги `<system-reminder>` самостійно.
+
+---
+
+## Як виглядає шаблон експерта?
+
+Кожен вбудований експерт (Code Expert / Research Expert) — це, по суті, блок `<system-reminder>`, що впроваджується у контекст Claude Code. Ваш користувацький експерт проходить через той самий конвеєр. Ось розбір шаблону **Research Expert**:
+
+```xml
+<system-reminder>
+[SCOPED INSTRUCTION] The following instructions are intended for the next 1–3
+interactions. Once the task is complete, these instructions should be gradually
+deprioritized and no longer influence subsequent interactions.
+
+Pre-requisite: Use `AskUserQuestion` to clarify the research scope, target
+audience, and deliverable format whenever the user's intent is ambiguous. Skip
+only if the intent is unambiguous.
+
+Leverage a multi-agent exploration mechanism to formulate an exceptionally
+detailed implementation plan.
+
+Instructions:
+1. Utilize the Agent tool to spawn parallel agents that simultaneously explore
+   various facets of the requirements:
+   - If necessary, deploy a preliminary investigator to conduct an initial
+     survey of industry-specific solutions using `webSearch`;
+   - If necessary, deploy a specialized investigator to research authoritative
+     sources—such as academic papers, news articles, and research reports—
+     using `webSearch`;
+   - Assign an agent to synthesize the target solution, while simultaneously
+     verifying the rigor and credibility of the gathered papers, news, and
+     research reports;
+   - If necessary, assign an agent to analyze competitor data to provide
+     supplementary analytical perspectives;
+   - If necessary, assign an agent to handle the implementation of a product
+     demo (generating outputs such as HTML, Markdown, etc.);
+   - If the task is sufficiently complex, you may assign additional teammates
+     to the roles defined above, or introduce other specialized roles; you are
+     permitted to schedule up to 5 teammates concurrently.
+
+2. Synthesize the findings from the aforementioned agents into a comprehensive,
+   step-by-step implementation plan.
+
+3. Utilize the Agent tool to spawn a set of parallel review agents; these
+   agents shall scrutinize the plan from multiple roles and perspectives to
+   identify any omitted steps and to propose reasonable additions or
+   optimizations.
+
+4. Consolidate the feedback received from the review agents, then invoke
+   `ExitPlanMode` to submit your final plan.
+
+5. Upon receiving the result from `ExitPlanMode`:
+   - If Approved: Proceed to execute the plan within this current session.
+   - If Rejected: Revise the plan based on the provided feedback, and then
+     invoke `ExitPlanMode` once again.
+   - If an Error Occurs: Do *not* follow the suggestions; prompt the user for
+     further instructions.
+
+Your final plan must include the following elements:
+- A clear summary of the proposed implementation strategy;
+- An ordered list of files to be created or modified, specifying the exact
+  changes required for each;
+- A step-by-step sequence for executing the implementation;
+- Identification of potential risks and corresponding mitigation strategies;
+- Creative ideation and suggestions for advanced enhancements;
+- If a product demo was generated, place the corresponding demo output in an
+  appropriate location and notify the user.
+</system-reminder>
+```
+
+---
+
+## Розбір за розділами
+
+### 1. Заголовок області `[SCOPED INSTRUCTION]` (обгортка — генерується автоматично)
+> The following instructions are intended for the next 1–3 interactions...
+
+Це повідомляє Claude Code: **ці інструкції активні лише для наступних 1–3 ходів**, потім згасають. Запобігає «витоку» «персони експерта» у подальші непов'язані діалоги.
+
+**Цей рядок генерується cc-viewer автоматично. Вам не потрібно його писати.**
+
+### 2. Початкове визначення завдання (**це те, що ви маєте переписати**)
+> Leverage a multi-agent exploration mechanism to formulate an exceptionally detailed implementation plan.
+
+Це «підмет-присудок-додаток» усього шаблону: **він повідомляє Claude Code позицію та мету**. Стандартна «багатоагентна розвідка + план реалізації» добре підходить для задач **програмної інженерії / планування**, але виглядає незграбно для багатьох інших галузей (рецензування контенту, аналіз даних, копірайтинг, маркетингові дослідження, аудит відповідності…).
+
+**Ми наполегливо рекомендуємо переписати цей рядок під вашу мету**, наприклад:
+
+- **Рецензент контенту**: "You are a senior content reviewer specializing in {domain}. Your goal is to identify factual inaccuracies, tone inconsistencies, and structural weaknesses in the provided material."
+- **Конкурентний аналітик**: "Conduct a rigorous competitive analysis for {product category}. Produce a comparison matrix, positioning insights, and strategic recommendations."
+- **Копірайтер**: "Generate multiple creative copy variants for {scenario}, each with distinct positioning, tone, and call-to-action strategy."
+
+### 3. Кроки робочого процесу (1–5 пунктів — **скоротіть або розширте залежно від складності**)
+
+Research Expert містить 5 кроків: **дослідження → синтез → рев'ю → надсилання плану → виконання**. Це забезпечує «паралельну багатоагентну роботу + перехресне рев'ю + затвердження плану» — три рівні строгості, придатні для задач високих ставок/широкого охоплення, але **надлишкові для легких**.
+
+- **Просте завдання** (одиночний пошук / маленьке виправлення): відмовтеся від диспетчеризації багатоагентності та рев'ю; просто «видайте відповідь» за один крок.
+- **Помірне завдання**: залиште «дослідження → синтез → рев'ю»; приберіть танець з ExitPlanMode; видайте результат напряму.
+- **Складне, дороге завдання** (великий рефакторинг, порівняння кількох варіантів, крос-доменне дослідження): збережіть усі 5 кроків, можливо додайте крок «модель ризиків» або «матриця порівняння варіантів».
+
+### 4. Підролі у кроці 1 (**адаптуйте під вашу галузь**)
+
+Research Expert перелічує 6 потенційних ролей (галузевий розвідник, академічний дослідник, синтезатор + перевіряючий факти, аналітик конкурентів, продюсер демо, розширюваний слот). **Перепишіть цей список під ваш сценарій**:
+
+- **Письмо**: "source collector + style analyst + fact checker"
+- **Аналіз даних**: "data-cleaning agent + statistical modeling agent + visualization agent"
+- **Аудит коду**: "static-analysis agent + dependency-chain auditor + threat modeler"
+
+### 5. Контрольний список підсумкового результату (**узгодьте з вашою реальною потребою**)
+
+> Your final plan must include the following elements: ...
+
+Оригінальний шаблон перелічує 6 елементів «плану реалізації». Ваш підсумковий результат може бути цілком іншим:
+
+- **Дослідницький звіт** → "Executive summary / Methodology / Key findings / Limitations / Action recommendations"
+- **Звіт про рев'ю** → "Issue list / Severity rating / Fix suggestions / Before-after examples"
+- **Матриця порівняння** → "Dimension definitions / Scoring rubric / Conclusions / Recommendation rationale"
+
+---
+
+## Поради щодо створення (TL;DR)
+
+1. **Збережіть обгортку**: рядок `<system-reminder>` + `[SCOPED INSTRUCTION]` додається cc-viewer — не повторюйте.
+2. **Перепишіть вступне речення**: вкажіть роль, мету та формат виводу в одному рядку.
+3. **Гнучко використовуйте робочий процес**: 1–2 кроки для легких завдань, повний 5-кроковий цикл — лише для складних.
+4. **Перепишіть підролі кроку 1**: стандартні (академічні статті / конкуренти / демо), ймовірно, не те, що вам потрібно.
+5. **Фінальний «контрольний список результату» — це ваша планка якості**: пропишіть структуру виводу — Claude Code суворо її дотримуватиметься.
+
+---
+
+## Рефакторений приклад: Competitive Analyst
+
+```
+You are a senior competitive intelligence analyst for {industry}. Your goal is to
+produce a decision-grade competitive landscape report for the product "{our product}".
+
+Instructions:
+1. Use the Agent tool to dispatch 3 parallel investigators:
+   - Market landscape agent: map the top 5–8 competitors with core positioning
+   - Feature matrix agent: compile a feature-by-feature comparison using
+     publicly available sources (webSearch)
+   - Pricing & GTM agent: analyze pricing models, distribution channels, and
+     go-to-market motions
+
+2. Synthesize the three streams into a unified competitive report.
+
+3. Dispatch one review agent to stress-test the report: challenge any
+   assumption lacking a cited source, flag outdated data (>12 months), and
+   propose one "non-obvious" insight.
+
+4. Deliver the final report with the following sections:
+   - TL;DR (3 bullets)
+   - Competitor positioning map
+   - Feature matrix (markdown table)
+   - Pricing & GTM table
+   - Top 3 strategic implications for our product
+   - Caveats & data gaps
+```
+
+Порівняно з оригінальним Research Expert: скорочено до 4 кроків, підролі зменшено з 6 до 3, список результатів повністю переписано як «розділи звіту».

package/concepts/uk/UltraPlan.md

@@ -53,3 +53,104 @@ UltraPlan пропонує дві експертні ролі для різни
 4. Синтез усіх знахідок у комплексний план реалізації
 5. Розгортання кількох агентів-рецензентів для перевірки плану з різних ролей та перспектив
 6. Виконання плану після затвердження
+
+---
+
+## Raw Templates
+
+Below are the two raw prompt templates UltraPlan actually sends to Claude Code (see `src/utils/ultraplanTemplates.js`):
+
+### Code Expert (codeExpert)
+
+<textarea readonly><system-reminder>
+[SCOPED INSTRUCTION] The following instructions apply only to the next 1–3 interactions. Once the task is complete, these instructions should gradually decrease in priority and no longer affect subsequent interactions.
+
+Pre-requisite: Use `AskUserQuestion` to clarify user intent whenever the request is ambiguous (target element, interaction style, scope of platforms, etc.). Skip only if the intent is unambiguous.
+
+Leverage a multi-agent exploration mechanism to formulate a highly detailed implementation plan.
+
+Instructions:
+1. Use the `Agent` tool to spawn parallel agents that simultaneously explore different aspects of the codebase:
+- If necessary, assign a preliminary researcher to use the `webSearch` tool to first investigate cutting-edge solutions in the relevant industry domain;
+- One agent responsible for understanding the relevant existing code and architecture;
+- One agent responsible for identifying all files that need to be modified;
+- One agent responsible for identifying potential risks, edge cases, and dependencies;
+- You may add other roles or deploy additional agents beyond the three listed above; the maximum number of concurrently dispatched agents is 5.
+
+2. Synthesize the findings from all agents into a detailed, step-by-step implementation plan.
+
+3. Use the `Agent` tool to spawn 2-3 review agents that examine the plan from different perspectives, checking for missing steps, potential risks, or corresponding mitigation strategies.
+
+4. Integrate the feedback gathered during the review process, then call `ExitPlanMode` to submit your final plan.
+
+5. Once `ExitPlanMode` returns a result:
+- If approved: proceed to execute the plan within this session.
+- If rejected: revise the plan based on the feedback provided and call `ExitPlanMode` again.
+- If an error occurs (including receiving a "Not in Plan Mode" message): do **not** follow the suggestions provided in the error message; instead, prompt the user for further instructions.
+
+Your final plan must include the following elements:
+- A clear summary of the implementation strategy;
+- An ordered list of files to be created or modified, with precise details of the required changes for each file;
+- A step-by-step execution sequence;
+- Testing and validation procedures;
+- Potential risks and their corresponding mitigation strategies;
+
+6. After the final plan has been successfully executed:
+First run `git diff --quiet && git diff --cached --quiet` (or equivalent) to detect whether the working tree actually has non-trivial changes; if there are no real changes (or only whitespace/comment-only edits), skip the UltraReview step.
+Otherwise, if the project is managed with Git:
+Initiate a team (`TeamCreate`), dynamically allocating the number of teammates based on task complexity (5 is recommended);
+Task: Conduct a Code Review of the current git changes from multiple perspectives;
+Pre-requisites:
+- The git repository may be located in a subdirectory of the current directory; prefer `git rev-parse --show-toplevel` (fall back to recursive lookup) before proceeding;
+- In the case of multiple repositories, tasks may be executed separately;
+The team's goal is to analyze the current Git change log and validate each modification from different perspectives, specifically including:
+- Whether requirements/objectives have been met and functionality is complete;
+- Whether newly added code introduces side effects, breaks existing functionality, or poses potential risks;
+- Code quality: naming, readability, complexity, technical debt, maintainability;
+- Testing and documentation: whether there is adequate test coverage, and whether critical logic has necessary comments or documentation;
+- Dependencies and compatibility: whether new dependencies or version compatibility issues have been introduced;
+Workflow:
+- Each teammate, according to their own role, covers the review dimensions one by one and independently outputs a report;
+- After consolidating the reports, perform a cross-review to identify conflicts or shared concerns;
+- Distill specific, actionable modification suggestions and annotate them with priority levels (P0/P1/P2/P3);
+- Upon completion, adopt P0 items, and selectively adopt P1 items when they are concrete and low-risk; defer P2/P3 to backlog;
+- After execution is complete, close the team (`TeamDelete`);
+</system-reminder></textarea>
+
+### Research Expert (researchExpert)
+
+<textarea readonly><system-reminder>
+[SCOPED INSTRUCTION] The following instructions are intended for the next 1–3 interactions. Once the task is complete, these instructions should be gradually deprioritized and no longer influence subsequent interactions.
+
+Pre-requisite: Use `AskUserQuestion` to clarify the research scope, target audience, and deliverable format whenever the user's intent is ambiguous. Skip only if the intent is unambiguous.
+
+Leverage a multi-agent exploration mechanism to formulate an exceptionally detailed implementation plan.
+
+Instructions:
+1. Utilize the Agent tool to spawn parallel agents that simultaneously explore various facets of the requirements:
+- If necessary, deploy a preliminary investigator to conduct an initial survey of industry-specific solutions using `webSearch`;
+- If necessary, deploy a specialized investigator to research authoritative sources—such as academic papers, news articles, and research reports—using `webSearch`;
+- Assign an agent to synthesize the target solution, while simultaneously verifying the rigor and credibility of the gathered papers, news, and research reports;
+- If necessary, assign an agent to analyze competitor data to provide supplementary analytical perspectives;
+- If necessary, assign an agent to handle the implementation of a product demo (generating outputs such as HTML, Markdown, etc.);
+- If the task is sufficiently complex, you may assign additional teammates to the roles defined above, or introduce other specialized roles; you are permitted to schedule up to 5 teammates concurrently.
+
+2. Synthesize the findings from the aforementioned agents into a comprehensive, step-by-step implementation plan.
+
+3. Utilize the Agent tool to spawn a set of parallel review agents; these agents shall scrutinize the plan from multiple roles and perspectives to identify any omitted steps and to propose reasonable additions or optimizations.
+
+4. Consolidate the feedback received from the review agents, then invoke `ExitPlanMode` to submit your final plan.
+
+5. Upon receiving the result from `ExitPlanMode`:
+- If Approved: Proceed to execute the plan within this current session.
+- If Rejected: Revise the plan based on the provided feedback, and then invoke `ExitPlanMode` once again.
+- If an Error Occurs (including the message "Not in Plan Mode"): Do *not* follow the suggestions provided by the error message; instead, prompt the user for further instructions.
+
+Your final plan must include the following elements:
+- A clear summary of the proposed implementation strategy;
+- An ordered list of files to be created or modified, specifying the exact changes required for each;
+- A step-by-step sequence for executing the implementation;
+- Identification of potential risks and corresponding mitigation strategies;
+- Creative ideation and suggestions for advanced enhancements;
+- If a product demo was generated, place the corresponding demo output in an appropriate location and notify the user.
+</system-reminder></textarea>

package/concepts/zh/CustomUltraplanExpert.md

@@ -17,6 +17,10 @@
 interactions. Once the task is complete, these instructions should be gradually
 deprioritized and no longer influence subsequent interactions.
 
+Pre-requisite: Use `AskUserQuestion` to clarify the research scope, target
+audience, and deliverable format whenever the user's intent is ambiguous. Skip
+only if the intent is unambiguous.
+
 Leverage a multi-agent exploration mechanism to formulate an exceptionally
 detailed implementation plan.
 

package/concepts/zh/UltraPlan.md

@@ -63,42 +63,58 @@ UltraPlan 提供两个专家角色，针对不同类型的任务：
 ### 代码专家（codeExpert）
 
 <textarea readonly><system-reminder>
-[SCOPED INSTRUCTION] The following instructions are intended for the next 1–3 interactions. Once the task is complete, these instructions should be gradually deprioritized and no longer influence subsequent interactions.
+[SCOPED INSTRUCTION] The following instructions apply only to the next 1–3 interactions. Once the task is complete, these instructions should gradually decrease in priority and no longer affect subsequent interactions.
+
+Pre-requisite: Use `AskUserQuestion` to clarify user intent whenever the request is ambiguous (target element, interaction style, scope of platforms, etc.). Skip only if the intent is unambiguous.
 
 Leverage a multi-agent exploration mechanism to formulate a highly detailed implementation plan.
 
 Instructions:
-1. Use the Agent tool to spawn parallel agents to simultaneously explore different aspects of the codebase:
-- If necessary, designate a preliminary investigator to use `webSearch` to first research advanced solutions within the relevant industry domain;
+1. Use the `Agent` tool to spawn parallel agents that simultaneously explore different aspects of the codebase:
+- If necessary, assign a preliminary researcher to use the `webSearch` tool to first investigate cutting-edge solutions in the relevant industry domain;
 - One agent responsible for understanding the relevant existing code and architecture;
-- One agent responsible for identifying all files requiring modification;
+- One agent responsible for identifying all files that need to be modified;
 - One agent responsible for identifying potential risks, edge cases, and dependencies;
-- You may add other roles or deploy additional agents beyond the three defined above; the maximum limit for concurrently scheduled agents is 5.
+- You may add other roles or deploy additional agents beyond the three listed above; the maximum number of concurrently dispatched agents is 5.
 
-2. Synthesize the findings from the aforementioned agents into a detailed, step-by-step implementation plan.
+2. Synthesize the findings from all agents into a detailed, step-by-step implementation plan.
 
-3. Use the Agent tool to spawn a review agent to scrutinize the plan from various perspectives, checking for any omitted steps, potential risks, or corresponding mitigation strategies.
+3. Use the `Agent` tool to spawn 2-3 review agents that examine the plan from different perspectives, checking for missing steps, potential risks, or corresponding mitigation strategies.
 
-4. Incorporate the feedback from the review process, then invoke `ExitPlanMode` to submit your final plan.
+4. Integrate the feedback gathered during the review process, then call `ExitPlanMode` to submit your final plan.
 
 5. Once `ExitPlanMode` returns a result:
-- If approved: Proceed to execute the plan within this session.
-- If rejected: Revise the plan based on the feedback provided, and invoke `ExitPlanMode` again.
-- If an error occurs (including the message "Not in Plan Mode"): Do *not* follow the suggestions provided by the error message; instead, prompt the user for further instructions.
+- If approved: proceed to execute the plan within this session.
+- If rejected: revise the plan based on the feedback provided and call `ExitPlanMode` again.
+- If an error occurs (including receiving a "Not in Plan Mode" message): do **not** follow the suggestions provided in the error message; instead, prompt the user for further instructions.
 
 Your final plan must include the following elements:
 - A clear summary of the implementation strategy;
-- An ordered list of files to be created or modified, specifying the exact changes required for each;
-- A step-by-step sequence for execution;
-- Testing and verification procedures;
-- Potential risks and their corresponding mitigation strategies.
-
-6. Upon the successful completion of the final plan's execution:
-If code changes have been made and the project is Git-based, invoke `TeamCreate` to assemble a "Code Review Team." The team's objective is to analyze the current Git change log and validate the modifications from various perspectives and roles, specifically to:
-- Confirm that the original requirements and objectives have been met;
-- Review any newly added code for potential side effects or regressions that might disrupt existing functionality;
-- Review the implemented code for any oversights or errors.
-Once the review report is generated, analyze it to formulate a set of recommended modifications; proceed to implement these recommended modifications by default.
+- An ordered list of files to be created or modified, with precise details of the required changes for each file;
+- A step-by-step execution sequence;
+- Testing and validation procedures;
+- Potential risks and their corresponding mitigation strategies;
+
+6. After the final plan has been successfully executed:
+First run `git diff --quiet && git diff --cached --quiet` (or equivalent) to detect whether the working tree actually has non-trivial changes; if there are no real changes (or only whitespace/comment-only edits), skip the UltraReview step.
+Otherwise, if the project is managed with Git:
+Initiate a team (`TeamCreate`), dynamically allocating the number of teammates based on task complexity (5 is recommended);
+Task: Conduct a Code Review of the current git changes from multiple perspectives;
+Pre-requisites:
+- The git repository may be located in a subdirectory of the current directory; prefer `git rev-parse --show-toplevel` (fall back to recursive lookup) before proceeding;
+- In the case of multiple repositories, tasks may be executed separately;
+The team's goal is to analyze the current Git change log and validate each modification from different perspectives, specifically including:
+- Whether requirements/objectives have been met and functionality is complete;
+- Whether newly added code introduces side effects, breaks existing functionality, or poses potential risks;
+- Code quality: naming, readability, complexity, technical debt, maintainability;
+- Testing and documentation: whether there is adequate test coverage, and whether critical logic has necessary comments or documentation;
+- Dependencies and compatibility: whether new dependencies or version compatibility issues have been introduced;
+Workflow:
+- Each teammate, according to their own role, covers the review dimensions one by one and independently outputs a report;
+- After consolidating the reports, perform a cross-review to identify conflicts or shared concerns;
+- Distill specific, actionable modification suggestions and annotate them with priority levels (P0/P1/P2/P3);
+- Upon completion, adopt P0 items, and selectively adopt P1 items when they are concrete and low-risk; defer P2/P3 to backlog;
+- After execution is complete, close the team (`TeamDelete`);
 </system-reminder></textarea>
 
 ### 调研专家（researchExpert）
@@ -106,6 +122,8 @@ Once the review report is generated, analyze it to formulate a set of recommende
 <textarea readonly><system-reminder>
 [SCOPED INSTRUCTION] The following instructions are intended for the next 1–3 interactions. Once the task is complete, these instructions should be gradually deprioritized and no longer influence subsequent interactions.
 
+Pre-requisite: Use `AskUserQuestion` to clarify the research scope, target audience, and deliverable format whenever the user's intent is ambiguous. Skip only if the intent is unambiguous.
+
 Leverage a multi-agent exploration mechanism to formulate an exceptionally detailed implementation plan.
 
 Instructions:

package/concepts/zh-TW/CustomUltraplanExpert.md

@@ -0,0 +1,162 @@
+# 自訂 UltraPlan 專家 — 使用說明
+
+## 兩個輸入框的作用
+
+- **專家名稱**：顯示在 UltraPlan 彈窗變體列裡的按鈕標題，30 字元以內。只是一個標籤，不會被傳送給 Claude Code。
+- **提示詞正文**：你寫給 Claude Code 的「角色指令」。傳送時會被**自動**用 `<system-reminder>...</system-reminder>` 標籤包裹，並在最前面加上一行 `[SCOPED INSTRUCTION]` 的作用域宣告。所以你**只需要寫正文**，不要自己加 `<system-reminder>` 標籤。
+
+---
+
+## 專家範本長什麼樣
+
+所有內建專家（Code Expert / Research Expert）本質上都是一段注入到 Claude Code 上下文裡的 `<system-reminder>`。你的自訂專家會走一模一樣的管線。下面以 **Research Expert** 為例拆解給你看：
+
+```xml
+<system-reminder>
+[SCOPED INSTRUCTION] The following instructions are intended for the next 1–3
+interactions. Once the task is complete, these instructions should be gradually
+deprioritized and no longer influence subsequent interactions.
+
+Pre-requisite: Use `AskUserQuestion` to clarify the research scope, target
+audience, and deliverable format whenever the user's intent is ambiguous. Skip
+only if the intent is unambiguous.
+
+Leverage a multi-agent exploration mechanism to formulate an exceptionally
+detailed implementation plan.
+
+Instructions:
+1. Utilize the Agent tool to spawn parallel agents that simultaneously explore
+   various facets of the requirements:
+   - If necessary, deploy a preliminary investigator to conduct an initial
+     survey of industry-specific solutions using `webSearch`;
+   - If necessary, deploy a specialized investigator to research authoritative
+     sources—such as academic papers, news articles, and research reports—
+     using `webSearch`;
+   - Assign an agent to synthesize the target solution, while simultaneously
+     verifying the rigor and credibility of the gathered papers, news, and
+     research reports;
+   - If necessary, assign an agent to analyze competitor data to provide
+     supplementary analytical perspectives;
+   - If necessary, assign an agent to handle the implementation of a product
+     demo (generating outputs such as HTML, Markdown, etc.);
+   - If the task is sufficiently complex, you may assign additional teammates
+     to the roles defined above, or introduce other specialized roles; you are
+     permitted to schedule up to 5 teammates concurrently.
+
+2. Synthesize the findings from the aforementioned agents into a comprehensive,
+   step-by-step implementation plan.
+
+3. Utilize the Agent tool to spawn a set of parallel review agents; these
+   agents shall scrutinize the plan from multiple roles and perspectives to
+   identify any omitted steps and to propose reasonable additions or
+   optimizations.
+
+4. Consolidate the feedback received from the review agents, then invoke
+   `ExitPlanMode` to submit your final plan.
+
+5. Upon receiving the result from `ExitPlanMode`:
+   - If Approved: Proceed to execute the plan within this current session.
+   - If Rejected: Revise the plan based on the provided feedback, and then
+     invoke `ExitPlanMode` once again.
+   - If an Error Occurs: Do *not* follow the suggestions; prompt the user for
+     further instructions.
+
+Your final plan must include the following elements:
+- A clear summary of the proposed implementation strategy;
+- An ordered list of files to be created or modified, specifying the exact
+  changes required for each;
+- A step-by-step sequence for executing the implementation;
+- Identification of potential risks and corresponding mitigation strategies;
+- Creative ideation and suggestions for advanced enhancements;
+- If a product demo was generated, place the corresponding demo output in an
+  appropriate location and notify the user.
+</system-reminder>
+```
+
+---
+
+## 逐段解說
+
+### 1. `[SCOPED INSTRUCTION]` 作用域宣告（外殼，自動產生）
+> The following instructions are intended for the next 1–3 interactions...
+
+這行告訴 Claude Code：這段指令**只在未來 1–3 輪對話中強制執行**，任務完成後就淡出，不再影響後續互動。避免「專家人格」汙染到任務之外的對話。
+
+**這行由 ccv 自動產生，你不需要寫。**
+
+### 2. 首段任務定義（**這段最值得你改**）
+> Leverage a multi-agent exploration mechanism to formulate an exceptionally detailed implementation plan.
+
+這是整段範本的「主謂賓」：**告訴 Claude Code 要以什麼姿態、為什麼目標工作**。預設的「多代理探索 + 實作計畫」適合**軟體研發／規劃型**任務，但對許多其他場景（內容審校、資料分析、文案創意、市場研究、合規稽核……）就不貼切。
+
+**建議你根據自己的目標改寫這一段**，比如：
+- **內容審校專家**："You are a senior content reviewer specializing in {領域}. Your goal is to identify factual inaccuracies, tone inconsistencies, and structural weaknesses in the provided material."
+- **競品分析專家**："Conduct a rigorous competitive analysis for {產品類別}. Produce a comparison matrix, positioning insights, and strategic recommendations."
+- **文案創意專家**："Generate multiple creative copy variants for {場景}, each with distinct positioning, tone, and call-to-action strategy."
+
+### 3. 工作流程步驟（1–5 項，**依任務複雜度適度精簡或擴充**）
+
+Research Expert 原範本是 5 步：**探索 → 彙整 → 審查 → 提交計畫 → 執行**。這套流程自帶「多代理並行 + 交叉審查 + 計畫審批」三層保障，適合跨度大、代價高的任務。但對許多輕量任務**過重了**，建議：
+
+- **任務簡單**（單點查詢／小修改）：刪掉多代理 agent 派遣、刪掉並行審查，直接 1 步「給出結論」。
+- **任務中等**：保留「探索 → 綜合 → 審查」三步，刪掉 ExitPlanMode 環節，直接產出結果。
+- **任務複雜、代價高**（大重構、多方案比較、跨多個領域研究）：保留 5 步，甚至增加「風險建模」或「方案比較矩陣」步驟。
+
+### 4. Step 1 的子角色清單（**依領域裁剪**）
+
+Research Expert 派了 6 種潛在角色（industry 調研員、學術研究員、綜合 + 核實、競品、Demo 產出、擴充角色）。**根據你的場景重新列清單**：
+- 寫作類：可能需要「資料蒐集員 + 風格分析員 + 事實查核員」
+- 資料分析：可能需要「資料清洗員 + 統計建模員 + 視覺化員」
+- 程式碼稽核：可能需要「靜態掃描員 + 相依鏈稽核員 + 威脅建模員」
+
+### 5. 最終產出清單（**對齊你的真正訴求**）
+
+> Your final plan must include the following elements: ...
+
+原範本要求的是「實作計畫」的 6 個要素，但你的任務產出可能是：
+- 一份研究報告 → 改為「摘要 / 方法論 / 核心發現 / 限制 / 行動建議」
+- 一份審校報告 → 改為「問題清單 / 嚴重度分級 / 修改建議 / 修改後示例」
+- 一份比較矩陣 → 改為「面向定義 / 評分標準 / 結論 / 優選理由」
+
+---
+
+## 撰寫建議（TL;DR）
+
+1. **保留外殼**（`<system-reminder>` 和 `[SCOPED INSTRUCTION]` 那行 ccv 自動加，你別重複寫）。
+2. **重點改首段**：用一句話明確角色、目標、產出形式。
+3. **流程依複雜度伸縮**：輕量任務 1–2 步就夠了，複雜任務才走 5 步的多代理 + 審查閉環。
+4. **Step 1 的子角色依領域重寫**：預設那些（academic papers / competitor / demo）大機率不是你要的。
+5. **最後的「產出清單」是 quality bar**：寫清楚你期望的輸出結構，Claude Code 會嚴格按這個產出結果。
+
+---
+
+## 一個改好的範例（競品分析專家）
+
+```
+You are a senior competitive intelligence analyst for {行業}. Your goal is to
+produce a decision-grade competitive landscape report for the product "{我方產品}".
+
+Instructions:
+1. Use the Agent tool to dispatch 3 parallel investigators:
+   - Market landscape agent: map the top 5–8 competitors with core positioning
+   - Feature matrix agent: compile a feature-by-feature comparison using
+     publicly available sources (webSearch)
+   - Pricing & GTM agent: analyze pricing models, distribution channels, and
+     go-to-market motions
+
+2. Synthesize the three streams into a unified competitive report.
+
+3. Dispatch one review agent to stress-test the report: challenge any
+   assumption lacking a cited source, flag outdated data (>12 months), and
+   propose one "non-obvious" insight.
+
+4. Deliver the final report with the following sections:
+   - TL;DR (3 bullets)
+   - Competitor positioning map
+   - Feature matrix (markdown table)
+   - Pricing & GTM table
+   - Top 3 strategic implications for our product
+   - Caveats & data gaps
+```
+
+這份相比 Research Expert 原版：精簡到 4 步，子角色從 6 個減到 3 個，產出清單完全重寫成「報告章節」。