email-whiz
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseEmail Whiz
Email Whiz
Gmail inbox management via MCP. Read current labels and filters before taking action.
基于MCP的Gmail收件箱管理工具,执行操作前请先读取当前标签和过滤器配置。
Canonical Vocabulary
标准术语表
| Term | Definition |
|---|---|
| 4D+N | DO / DELEGATE / DEFER / REFERENCE / NOISE — five triage buckets |
| inbox zero | State where inbox contains only items requiring action today |
| routine | Repeatable time-boxed workflow (daily 5 min, weekly 15 min) |
| streak | Consecutive days achieving inbox zero; stored in progress file |
| auto-rule | Gmail filter created from behavioral pattern analysis |
| filter | Gmail rule that matches incoming email and applies actions |
| VIP | Sender who receives consistent replies and high-priority treatment |
| noise | Automated or low-value email; candidate for filtering/deletion |
| engagement | Reply rate to a sender: replies sent ÷ emails received |
| bankruptcy | Inbox with 500+ unread or growing backlog — triggers staged recovery |
| confidence | HIGH >80% / MEDIUM 50-80% / LOW <50% — governs auto-rule actions |
| batch op | Single |
| 术语 | 定义 |
|---|---|
| 4D+N | 执行/委派/延后/参考/垃圾 — 五个邮件分类桶 |
| inbox zero | 收件箱仅留存当日需处理事项的状态 |
| routine | 可重复的限时工作流(每日5分钟,每周15分钟) |
| streak | 连续达成收件箱清零的天数,存储在进度文件中 |
| auto-rule | 基于行为模式分析创建的Gmail过滤器 |
| filter | 匹配入站邮件并执行对应操作的Gmail规则 |
| VIP | 可获得稳定回复和高优先级处理的发件人 |
| noise | 自动化发送或低价值邮件,适合过滤/删除 |
| engagement | 对某发件人的回复率:发送回复数 ÷ 接收邮件数 |
| bankruptcy | 收件箱有500+未读邮件或积压量持续增长,触发分阶段恢复流程 |
| confidence | 高 >80% / 中 50-80% / 低 <50% — 管控自动规则的执行逻辑 |
| batch op | 单次 |
Dispatch
指令调度
| Action |
|---|---|
| Categorize inbox with 4D+N, batch process by category |
| Daily/weekly routine, progress tracking, bankruptcy detection |
| Pattern-detected filter suggestions with confidence scoring |
| Behavioral analysis → auto-create Gmail filters from templates |
| Volume trends, response times, sender frequency, timing patterns |
| Subscription audit + unsubscribe plan |
| Taxonomy analysis, merge/rename/delete recommendations |
| Natural language → Gmail search query |
| VIP + noise identification |
| Recent important email summary |
| Archive candidates, duplicate detection, batch archiving |
| Full inbox health report with score and action plan |
| (empty) | Show this mode menu |
Classification gate for ambiguous input: If could match multiple modes, ask which the user wants before proceeding.
$ARGUMENTS | 动作 |
|---|---|
| 用4D+N框架对收件箱分类,按类别批量处理 |
| 每日/每周例行流程,进度追踪,积压预警检测 |
| 基于模式识别生成带置信度评分的过滤器建议 |
| 基于行为分析从模板自动创建Gmail过滤器 |
| 数量趋势、响应时间、发件人频率、时间模式分析 |
| 订阅审计 + 退订计划 |
| 分类体系分析,提供合并/重命名/删除建议 |
| 自然语言转Gmail搜索查询 |
| VIP和垃圾发件人识别 |
| 近期重要邮件摘要 |
| 识别可归档内容,重复检测,批量归档 |
| 完整的收件箱健康度报告,含评分和行动计划 |
| (空) | 显示此模式菜单 |
模糊输入分类规则: 如果匹配多个模式,先询问用户需求再继续执行。
$ARGUMENTSHybrid Mode Protocol
混合模式协议
Read operations — execute immediately:
- ,
gmail_search_emails,gmail_get_emailgmail_list_emails - ,
gmail_get_labels,gmail_get_filtersgmail_get_filter
Write operations — confirm before executing:
- Single: ,
gmail_modify_email,gmail_delete_emailgmail_create_label - Batch: ,
gmail_batch_modify_emailsgmail_batch_delete_emails - Labels: ,
gmail_update_label,gmail_delete_labelgmail_get_or_create_label - Filters: ,
gmail_create_filter,gmail_create_filter_from_templategmail_delete_filter
Confirmation format — always show before write ops:
Action: {description}
Scope: {count} email(s) / label(s) / filter(s)
Reversible: {Yes | NO}
[Sample: first 3 subjects for batch ops]Use for full confirmation and report formats.
templates.md读操作 — 立即执行:
- ,
gmail_search_emails,gmail_get_emailgmail_list_emails - ,
gmail_get_labels,gmail_get_filtersgmail_get_filter
写操作 — 执行前需确认:
- 单条操作:,
gmail_modify_email,gmail_delete_emailgmail_create_label - 批量操作:,
gmail_batch_modify_emailsgmail_batch_delete_emails - 标签操作:,
gmail_update_label,gmail_delete_labelgmail_get_or_create_label - 过滤器操作:,
gmail_create_filter,gmail_create_filter_from_templategmail_delete_filter
确认格式 — 写操作前必须展示:
Action: {description}
Scope: {count} email(s) / label(s) / filter(s)
Reversible: {Yes | NO}
[Sample: first 3 subjects for batch ops]完整的确认和报告格式请参考。
templates.mdPhase 0: Discovery (Run First)
阶段0:上下文发现(优先执行)
Before any workflow, establish context:
1. gmail_get_labels → store label names + IDs
2. gmail_get_filters → store existing filter criteria
3. gmail_search_emails → query: "in:inbox" (maxResults: 20)
note unread count, top senders, label patternsUse discovered label names throughout — never invent names that conflict with existing taxonomy.
执行任何工作流前,先确认上下文:
1. gmail_get_labels → 存储标签名称+ID
2. gmail_get_filters → 存储现有过滤器规则
3. gmail_search_emails → 查询: "in:inbox" (maxResults: 20)
记录未读数量、顶级发件人、标签使用模式全程使用已发现的标签名称,绝对不要创建和现有分类体系冲突的名称。
Mode: Triage
模式:邮件分类
Categorize inbox using 4D+N. Reference: .
triage-framework.md1. Search: is:unread (maxResults: 50)
2. For each email: classify as DO / DELEGATE / DEFER / REFERENCE / NOISE
3. Collect IDs by category
4. Batch process:
- NOISE ids: gmail_batch_modify_emails (removeLabelIds: [INBOX])
- REFERENCE ids: gmail_batch_modify_emails (add _reference, remove INBOX)
- DEFER ids: gmail_batch_modify_emails (add _deferred, remove INBOX)
5. Report with triage-summary template
6. Present DO items for user actionAfter triage: log session to inbox-zero progress file if running as part of daily routine.
使用4D+N框架对收件箱分类,参考文档:。
triage-framework.md1. 搜索: is:unread (maxResults: 50)
2. 对每封邮件分类为:执行/委派/延后/参考/垃圾
3. 按分类收集邮件ID
4. 批量处理:
- 垃圾邮件ID: gmail_batch_modify_emails (removeLabelIds: [INBOX])
- 参考类邮件ID: gmail_batch_modify_emails (add _reference, remove INBOX)
- 延后类邮件ID: gmail_batch_modify_emails (add _deferred, remove INBOX)
5. 使用分类摘要模板生成报告
6. 展示待执行类邮件供用户操作分类完成后:如果是每日例行流程的一部分,将会话记录写入收件箱清零进度文件。
Mode: Inbox Zero
模式:收件箱清零
Structured habit system for achieving and maintaining inbox zero. Reference: .
inbox-zero-system.md实现和维持收件箱清零的结构化习惯体系,参考文档:。
inbox-zero-system.mdDaily Routine (5 min)
每日例行流程(5分钟)
1. Search: is:unread newer_than:1d → identify P0/P1
2. Run quick triage pass (4D+N)
3. Batch archive NOISE + REFERENCE
4. Report: inbox count before → after
5. Update ~/.claude/email-whiz/inbox-zero-progress.json
- achieved_zero: inbox_count == 0
- increment streak if zero1. 搜索: is:unread newer_than:1d → 识别P0/P1优先级邮件
2. 快速执行4D+N分类
3. 批量归档垃圾+参考类邮件
4. 报告:收件箱数量 处理前 → 处理后
5. 更新 ~/.claude/email-whiz/inbox-zero-progress.json
- achieved_zero: 收件箱数量 == 0
- 如果达成清零则增加连续打卡天数Weekly Review (15 min)
每周回顾(15分钟)
1. Clear deferred: label:_deferred older_than:7d
2. Filter effectiveness: inbox noise still arriving? → new filter candidates
3. Newsletter sweep: label:_reading/newsletters is:unread older_than:14d
4. Update weekly_reviews entry in progress fileBankruptcy detection: If count > 500 or growing 3+ consecutive days, present recovery options (Daily Blitz / Staged Approach / Nuclear Reset). Always confirm before bulk archiving. See for full protocols.
is:unreadinbox-zero-system.mdProgress report format:
📊 Streak: {n} days | Best: {n} | Last zero: {date}
Inbox: {before} → {after}1. 清理延后邮件:label:_deferred older_than:7d
2. 过滤器有效性检查:是否仍有垃圾邮件进入收件箱?→ 生成新过滤器候选
3. 新闻通讯清理:label:_reading/newsletters is:unread older_than:14d
4. 更新进度文件中的每周回顾条目积压预警检测: 如果数量>500或连续3天持续增长,展示恢复选项(每日清理/分阶段处理/重置)。批量归档前必须获得用户确认,完整流程参考。
is:unreadinbox-zero-system.md进度报告格式:
📊 连续打卡: {n} 天 | 最高记录: {n} | 上次清零: {date}
收件箱: {before} → {after}Mode: Filters
模式:过滤器管理
Pattern-detected filter suggestions. Reference: .
filter-patterns.md1. Search recent inbox (maxResults: 100, last 30d)
2. Cluster by sender domain + subject patterns
3. Check gmail_get_filters for existing coverage
4. Score each candidate:
HIGH (>80%): volume ≥20, engagement <5% → offer to create
MEDIUM(50-80%): volume ≥10, engagement <15% → suggest
LOW (<50%): borderline → show only
5. Present with filter-suggestions template
6. On approval: gmail_create_filter or gmail_create_filter_from_templateCheck for conflicts before creating: scan existing filter criteria for overlaps.
基于模式识别生成过滤器建议,参考文档:。
filter-patterns.md1. 搜索近期收件箱邮件(maxResults: 100, 近30天)
2. 按发件人域名+主题模式聚类
3. 调用gmail_get_filters检查现有过滤器覆盖情况
4. 为每个候选过滤器评分:
高 (>80%): 数量≥20, 回复率<5% → 提供创建选项
中(50-80%): 数量≥10, 回复率<15% → 给出建议
低 (<50%): 边界情况 → 仅展示
5. 使用过滤器建议模板展示结果
6. 获得批准后:执行gmail_create_filter或gmail_create_filter_from_template创建前检查冲突:扫描现有过滤器规则是否存在重叠。
Mode: Auto-Rules
模式:自动规则
Deep behavioral analysis → automated filter creation. Reference: § Auto-Rule Detection.
filter-patterns.md深度行为分析→自动化过滤器创建,参考文档: § 自动规则检测。
filter-patterns.mdAnalysis Phase
分析阶段
1. Search: in:inbox after:{90d_ago} (maxResults: 500)
2. Sender cluster analysis:
- Compute volume + engagement per sender
- Identify: HIGH confidence (auto-create), MEDIUM (review), LOW (skip)
3. Subject pattern mining:
- Find repeating prefixes: [JIRA], PR #, Invoice #
- Score by occurrence count
4. Check conflict with existing filters1. 搜索: in:inbox after:{90天前} (maxResults: 500)
2. 发件人聚类分析:
- 计算每个发件人的邮件数量+回复率
- 识别:高置信度(自动创建)、中置信度(需审核)、低置信度(跳过)
3. 主题模式挖掘:
- 识别重复前缀:[JIRA], PR #, Invoice #
- 按出现次数评分
4. 检查与现有过滤器的冲突Creation Phase
创建阶段
5. Present auto-rules report (see templates.md)
6. HIGH confidence: gmail_create_filter_from_template with confirmation
MEDIUM confidence: gmail_create_filter after individual review
7. Schedule learning loop re-analysis after 2 weeks5. 展示自动规则报告(参考templates.md)
6. 高置信度:确认后调用gmail_create_filter_from_template
中置信度:单独审核后调用gmail_create_filter
7. 安排2周后重新执行学习循环分析Mode: Analytics
模式:数据分析
Email communication metrics and trends. Reference: .
analytics-guide.md1. Volume: count in:inbox per week (4-week trend)
- Compare week-over-week; classify growing/stable/declining
2. Inbox reach rate: direct inbox ÷ total received
- Good: <30% | Needs filters: >60%
3. Response health:
- is:starred older_than:2d → overdue
- Reply rate: in:sent subject:Re: ÷ in:inbox (30d sample)
4. Sender frequency:
- Top 20 senders by volume + engagement matrix
- Classify: VIP / Active / Passive / Noise
5. Time-of-day: extract hour from date fields in sample
6. Label distribution: activity per label, stale detection (no messages 90d+)
7. Present full analytics report (see templates.md)邮件沟通指标和趋势分析,参考文档:。
analytics-guide.md1. 邮件数量:每周收件箱数量(4周趋势)
- 周同比对比,分类为增长/稳定/下降
2. 收件箱到达率:直接进入收件箱邮件 ÷ 总接收邮件
- 优秀:<30% | 需要优化过滤器:>60%
3. 响应健康度:
- 已标星且超过2天的邮件 → 逾期
- 回复率:已发送主题含Re:的邮件 ÷ 收件箱邮件(30天样本)
4. 发件人频率:
- 按数量+回复率矩阵排序前20名发件人
- 分类:VIP / 活跃 / 被动 / 垃圾
5. 发送时段:从样本邮件的日期字段提取小时分布
6. 标签分布:每个标签的活跃度,识别闲置标签(90天以上无邮件)
7. 展示完整分析报告(参考templates.md)Mode: Newsletters
模式:新闻通讯管理
Newsletter subscription audit + unsubscribe plan.
1. Search: list:* OR subject:(newsletter OR digest OR "weekly roundup")
2. For each subscription: estimate frequency + read rate (recency heuristic)
3. Classify: UNSUBSCRIBE (low read rate) / KEEP / FILTER (keep but skip inbox)
4. Present newsletter-audit report
5. On approval: open unsubscribe links (provide URL) or create filterReference: § Unsubscribe Strategies.
workflows.md新闻通讯订阅审计+退订计划。
1. 搜索: list:* OR subject:(newsletter OR digest OR "weekly roundup")
2. 对每个订阅:估算发送频率+阅读率(基于时效性启发式规则)
3. 分类:退订(低阅读率)/ 保留 / 过滤(保留但不进入收件箱)
4. 展示新闻通讯审计报告
5. 获得批准后:打开退订链接(提供URL)或创建过滤器参考文档: § 退订策略。
workflows.mdMode: Labels
模式:标签管理
Label taxonomy analysis and cleanup.
1. gmail_get_labels → list all with counts
2. Find issues:
- Duplicates/synonyms (e.g., "work" + "Work" + "work-email")
- Stale labels (no messages in 90d)
- Flat structure that should be hierarchical
3. Suggest: MERGE / RENAME / DELETE / RESTRUCTURE
4. For merges: gmail_batch_modify_emails to migrate, then gmail_delete_label
5. Use gmail_update_label for renames
6. Present label-analysis reportReference: § Label Hierarchy Best Practices.
workflows.md标签分类体系分析和清理。
1. gmail_get_labels → 列出所有标签及对应邮件数量
2. 发现问题:
- 重复/同义词(例如 "work" + "Work" + "work-email")
- 闲置标签(90天无邮件)
- 可优化为层级结构的扁平结构
3. 给出建议:合并 / 重命名 / 删除 / 重构
4. 合并操作:调用gmail_batch_modify_emails迁移邮件,然后调用gmail_delete_label
5. 重命名操作调用gmail_update_label
6. 展示标签分析报告参考文档: § 标签层级最佳实践。
workflows.mdMode: Search
模式:搜索
Natural language → Gmail search query.
Input: user's goal in plain language
Output: optimized Gmail search query with explanation
Steps:
1. Parse intent: who / what / when / status
2. Map to operators: from:, subject:, after:, is:unread, label:, etc.
3. Add negations for common noise
4. Show query + explanation + expected matches
5. Offer to run the searchReference: § Gmail Search Operators.
filter-patterns.md自然语言→Gmail搜索查询。
输入:用户用自然语言描述的搜索目标
输出:优化后的Gmail搜索查询及解释
步骤:
1. 解析意图:谁 / 什么 / 何时 / 状态
2. 映射到操作符:from:, subject:, after:, is:unread, label: 等
3. 添加常见垃圾邮件的排除规则
4. 展示查询语句+解释+预期匹配结果
5. 提供执行搜索的选项参考文档: § Gmail搜索操作符。
filter-patterns.mdMode: Senders
模式:发件人管理
VIP and noise sender identification.
1. Sample last 200 inbox emails
2. Compute per sender: volume + engagement + recency
3. VIP candidates: reply rate >50% OR consistently starred
4. Noise candidates: volume >10 AND engagement 0%
5. Present sender report with recommended actions
6. On approval: create VIP filter (always important) or noise filterReference: § Sender Shortcuts, § VIP Management.
triage-framework.mdworkflows.mdVIP和垃圾发件人识别。
1. 采样最近200封收件箱邮件
2. 计算每个发件人的指标:邮件数量 + 回复率 + 时效性
3. VIP候选:回复率>50% 或 consistently starred
4. 垃圾发件人候选:邮件数量>10 且 回复率0%
5. 展示发件人报告及推荐操作
6. 获得批准后:创建VIP过滤器(始终标记为重要)或垃圾过滤器参考文档: § 发件人快捷方式, § VIP管理。
triage-framework.mdworkflows.mdMode: Digest
模式:摘要生成
Recent important email summary.
1. Search: is:important newer_than:3d (or is:starred OR is:unread is:important)
2. For each: extract sender, subject, 1-line summary
3. Categorize: RESPOND / FYI / READING
4. Present digest report
5. Offer: "open N" | "archive fyi"Reference: § Digest.
templates.md近期重要邮件摘要。
1. 搜索: is:important newer_than:3d (或 is:starred OR is:unread is:important)
2. 提取每封邮件的发件人、主题、1行摘要
3. 分类:需回复 / 仅供参考 / 待阅读
4. 展示摘要报告
5. 提供选项:"打开第N封" | "归档参考类邮件"参考文档: § 摘要。
templates.mdMode: Cleanup
模式:清理
Archive candidates and batch archiving.
1. Detect stale inbox items:
- Read receipts/confirmations: subject:(receipt OR confirmation) older_than:30d
- Old newsletters: label:_reading/newsletters older_than:30d
- Expired deferrals: label:_deferred older_than:14d
- Large threads fully read: is:read has:nouserlabels older_than:60d
2. Show batch-operation-confirmation with samples
3. On approval: gmail_batch_modify_emails (remove INBOX / add archive label)
4. Report: inbox count before → afterNever use in cleanup — archive only.
Reference: § Batch Operations.
gmail_batch_delete_emailsworkflows.md识别可归档内容和批量归档。
1. 检测收件箱闲置内容:
- 已读收据/确认函:subject:(receipt OR confirmation) older_than:30d
- 旧新闻通讯:label:_reading/newsletters older_than:30d
- 过期延后邮件:label:_deferred older_than:14d
- 已读无标签的长线程:is:read has:nouserlabels older_than:60d
2. 展示批量操作确认信息及样本
3. 获得批准后:调用gmail_batch_modify_emails(移除收件箱标签 / 添加归档标签)
4. 报告:收件箱数量 处理前 → 处理后清理模式下绝对不要使用,仅执行归档操作。
参考文档: § 批量操作。
gmail_batch_delete_emailsworkflows.mdMode: Audit
模式:审计
Full inbox health report.
1. Run Phase 0 discovery (labels, filters, sample)
2. Collect metrics:
- Inbox count + unread %
- Filter count + estimated coverage
- Label count + stale count
- Newsletter volume
- VIP / noise distribution
3. Score: 0-100 based on inbox reach rate, filter coverage, label health
4. Identify top 3 quick wins (high impact, low effort)
5. Generate week-by-week action plan
6. Present full-audit reportReference: , § Full Audit.
analytics-guide.mdtemplates.md完整的收件箱健康度报告。
1. 执行阶段0上下文发现(标签、过滤器、样本)
2. 收集指标:
- 收件箱总数 + 未读占比
- 过滤器数量 + 预估覆盖率
- 标签数量 + 闲置标签数量
- 新闻通讯数量
- VIP / 垃圾发件人分布
3. 评分:基于收件箱到达率、过滤器覆盖率、标签健康度给出0-100分
4. 识别Top3快速优化项(高收益低投入)
5. 生成逐周行动计划
6. 展示完整审计报告参考文档:, § 完整审计。
analytics-guide.mdtemplates.mdError Handling
错误处理
| Scenario | Response |
|---|---|
| MCP connection failure | Report error, provide manual Gmail URL as fallback |
| Rate limit (429) | Reduce batch size to 25, wait 60s, retry once |
| Message not found (404) | Skip and continue; note in report |
| Partial batch failure | Report succeeded/failed counts; suggest retry for failures |
| Label ID mismatch | Re-fetch labels via |
| 场景 | 响应 |
|---|---|
| MCP连接失败 | 报告错误,提供手动Gmail链接作为备选方案 |
| 限流(429) | 把批量大小降低到25,等待60秒,重试1次 |
| 消息未找到(404) | 跳过继续执行,在报告中标记 |
| 部分批量操作失败 | 报告成功/失败数量,建议对失败项重试 |
| 标签ID不匹配 | 重新调用 |
Scope Boundaries
适用范围边界
IS for: Gmail triage, inbox zero habits, filter creation, auto-rules, email analytics, newsletter management, label organization, search queries, sender analysis, digest, cleanup, inbox health audit.
NOT for: Composing or sending outbound emails, calendar scheduling, Google Drive, non-Gmail accounts, real-time email monitoring, CRM workflows.
适用场景: Gmail邮件分类、收件箱清零习惯养成、过滤器创建、自动规则、邮件数据分析、新闻通讯管理、标签整理、搜索查询生成、发件人分析、摘要生成、清理、收件箱健康度审计。
不适用场景: 撰写或发送外发邮件、日历日程安排、Google Drive操作、非Gmail账户、实时邮件监控、CRM工作流。
Reference File Index
参考文件索引
| File | Content | Load When |
|---|---|---|
| 5D+N decision trees, sender/subject signals, batch processing | triage, inbox-zero |
| Gmail operators, filter templates, auto-rule algorithms, learning loop | filters, auto-rules, search |
| Bankruptcy, VIP, batch ops, label hierarchy, unsubscribe, security | cleanup, labels, senders, newsletters |
| All confirmation, report, and execution result templates | Every mode |
| Daily/weekly routines, progress schema, bankruptcy recovery | inbox-zero |
| Volume, response time, sender frequency, timing methodology | analytics, audit |
| All 19 Gmail MCP tool signatures, parameters, system limits | When selecting tools |
Scripts:
| Script | Purpose | Run When |
|---|---|---|
| Save inbox count snapshot | After each inbox-zero check |
| Compute inbox trend (growing/stable/declining) | During analytics mode |
| List recent snapshots | When reviewing progress |
| 文件 | 内容 | 加载时机 |
|---|---|---|
| 5D+N决策树、发件人/主题信号、批量处理 | 分类、收件箱清零模式 |
| Gmail操作符、过滤器模板、自动规则算法、学习循环 | 过滤器、自动规则、搜索模式 |
| 积压恢复、VIP管理、批量操作、标签层级、退订、安全 | 清理、标签、发件人、新闻通讯模式 |
| 所有确认、报告、执行结果模板 | 所有模式 |
| 每日/每周例行流程、进度结构、积压恢复 | 收件箱清零模式 |
| 数量、响应时间、发件人频率、时间分析方法 | 数据分析、审计模式 |
| 所有19个Gmail MCP工具签名、参数、系统限制 | 选择工具时 |
脚本:
| 脚本 | 用途 | 运行时机 |
|---|---|---|
| 保存收件箱数量快照 | 每次收件箱清零检查后 |
| 计算收件箱趋势(增长/稳定/下降) | 数据分析模式期间 |
| 列出近期快照 | 进度回顾时 |
Critical Rules
核心规则
- Run Phase 0 discovery before any workflow — never assume label/filter state
- Confirm all write operations — show scope, count, and reversibility before calling
- Batch similar operations — never loop when
gmail_modify_emailappliesgmail_batch_modify_emails - Preserve existing taxonomy — suggest improvements, never silently rename
- Check filter conflicts before creating — call first
gmail_get_filters - Include confidence scores on all filter/rule suggestions
- Provide rollback guidance for every bulk operation
- Handle MCP failures gracefully — surface manual fallback steps
- Surface unsubscribe links — make newsletter cleanup actionable
- Prioritize quick wins — high impact, low effort first in all reports
- Update inbox-zero progress after every routine — write streak data to
~/.claude/email-whiz/inbox-zero-progress.json - Never use in cleanup mode — archive only; deletion requires explicit user intent
gmail_batch_delete_emails
- 任何工作流执行前先运行阶段0上下文发现,绝对不要假设标签/过滤器状态
- 所有写操作必须确认,调用前展示范围、数量、可恢复性
- 批量执行相似操作,可使用时绝对不要循环调用
gmail_batch_modify_emailsgmail_modify_email - 保留现有分类体系,仅给出改进建议,绝对不要静默重命名
- 创建过滤器前检查冲突,先调用
gmail_get_filters - 所有过滤器/规则建议必须附带置信度评分
- 每个批量操作都提供回滚指导
- 优雅处理MCP失败,给出手动备选步骤
- 展示退订链接,让新闻通讯清理可执行
- 优先展示快速优化项,所有报告优先展示高收益低投入的内容
- 每次例行流程后更新收件箱清零进度,将连续打卡数据写入
~/.claude/email-whiz/inbox-zero-progress.json - 清理模式下绝对不要使用,仅执行归档;删除操作需要用户明确授权
gmail_batch_delete_emails