OpenClaw Implementation Prompts
OpenClaw 实现提示词
Each prompt below is a self-contained brief you can hand to an AI coding assistant (or use as a project spec) to build that use case from scratch. Adapt the specific services to whatever you already use — the patterns are what matter.
下面的每个提示词都是一个独立的简报,你可以交给 AI 编程助手(或用作项目规范)来从头构建该用例。将具体服务调整为你已经在使用的任何服务——模式才是关键。
1) Personal CRM Intelligence
个人 CRM 智能
Build me a personal CRM system that automatically tracks everyone I interact with, with smart filtering so it only adds real people — not newsletters, bots, or cold outreach.
为我构建一个个人 CRM 系统,自动跟踪我与之互动的每个人,并通过智能过滤确保它只添加真实的人——而不是通讯邮件、机器人或陌生推销。
Data sources:
- Connect to my email (Gmail API or IMAP) and scan the last 60 days of messages.
- Connect to my calendar (Google Calendar API) and scan the last 60 days of events.
- Run this ingestion on a daily cron schedule.
数据源:
- 连接我的邮箱(Gmail API 或 IMAP)并扫描过去 60 天的邮件。
- 连接我的日历(Google Calendar API)并扫描过去 60 天的事件。
- 每天通过定时任务运行此数据摄取。
Contact extraction from email:
- Extract sender/recipient email addresses and names from messages.
- Estimate the number of exchanges (back-and-forth threads, not just raw message count): Math.min(Math.floor(totalMessages / 2), threadCount).
- Collect sample subject lines and message snippets for classification.
从邮件中提取联系人:
- 从邮件中提取发件人/收件人的邮箱地址和姓名。
- 估算交流次数(来回对话线程,不只是原始邮件数量):Math.min(Math.floor(totalMessages / 2), threadCount)。
- 收集样本主题行和邮件片段用于分类。
Contact extraction from calendar:
- Only include meetings with 1-10 attendees (skip large all-hands).
- Only include meetings at least 15 minutes long (skip quick check-ins that are really just reminders).
- Extract attendee names, emails, and the meeting title.
从日历中提取联系人:
- 只包含 1-10 名参会者的会议(跳过大型的全员会议)。
- 只包含至少 15 分钟的会议(跳过实际上是提醒的快速签到)。
- 提取参会者姓名、邮箱和会议标题。
Filtering — this is critical. Most contacts from email are noise. Use a two-stage filter:
过滤——这是关键。邮件中的大多数联系人都是噪音。使用两级过滤器:
Stage 1 — Hard filters (always reject):
- My own email addresses and domains.
- Emails from family or personal contacts I've explicitly excluded (configurable list).
- Contacts already in the CRM or previously rejected.
- Generic role-based inboxes: info@, team@, partnerships@, collabs@, noreply@.
- Marketing/transactional domains matching patterns like: noreply@, tx., cx., mail., email. prefixes.
第一阶段——硬过滤(总是拒绝):
- 我自己的邮箱地址和域名。
- 来自我已明确排除的家人或个人联系人的邮件(可配置列表)。
- 已在 CRM 中或之前被拒绝的联系人。
- 通用角色邮箱:info@、team@、partnerships@、collabs@、noreply@。
- 匹配以下模式的营销/交易域名:noreply@、tx.、cx.、mail.、email. 前缀。
Stage 2 — AI classification (use a fast, cheap LLM like Gemini Flash or Haiku):
Send the candidate's name, email, exchange count, and sample subject lines/snippets to an LLM with these rules:
- REJECT clearly automated or notification-only senders.
- REJECT if all sample subjects look like newsletters, digests, or automated reports ("weekly roundup", "monthly update", "AI news").
- REJECT cold outreach with low engagement — if exchanges are low relative to total emails, it's one-way pitching.
- REJECT if snippets show repetitive promotional content (product launches, feature announcements, affiliate reports).
- APPROVE only if it looks like a real person with genuine two-way interaction or a meaningful business relationship.
- Higher confidence for real back-and-forth conversations with varied, substantive topics.
第二阶段——AI 分类(使用快速、便宜的 LLM,如 Gemini Flash 或 Haiku):
将候选人的姓名、邮箱、交流次数和样本主题行/片段发送给 LLM,遵循以下规则:
- 拒绝明显自动化的或仅通知的发件人。
- 如果所有样本主题看起来像通讯邮件、摘要或自动报告("weekly roundup"、"monthly update"、"AI news"),则拒绝。
- 拒绝低互动的陌生推销——如果相对于总邮件数交流次数很低,说明是单向推销。
- 如果片段显示重复的推广内容(产品发布、功能公告、联盟报告),则拒绝。
- 只有看起来是真实的人且有真正的双向互动或有意义的业务关系时才批准。
- 对于涉及多样、实质性话题的真正来回对话,置信度更高。
Contact scoring (used for ranking, not filtering):
- Base score: 50
- +5 per email exchange (max +20)
- +3 per meeting (max +15)
- +15 if their title matches preferred titles (CEO, Founder, VP, Head of, Engineer, Partner, etc.)
- +10 if they appeared in small meetings (≤3 attendees)
+10 if last interaction was within 7 days, +5 if within 30 days
+25 bonus if the person appears in both email AND calendar (stronger signal)
+10 if they have a recognizable role, +5 if they have a company
联系人评分(用于排序,而非过滤):
基础分:50
每次邮件交流 +5(最高 +20)
每次会议 +3(最高 +15)
如果其职位匹配首选职位(CEO、Founder、VP、Head of、Engineer、Partner 等)+15
如果他们出现在小型会议中(≤3 名参会者)+10
如果最后一次互动在 7 天内 +10,30 天内 +5
如果该人同时出现在邮件和日历中(更强的信号)+25 奖励
如果有可识别的职位 +10,如果有公司 +5
For each approved contact, store:
Name, email(s), company, role/context
Interaction timeline with dates
Last-touch timestamp (auto-updated)
Contact score
Tags or categories
对于每个批准的联系人,存储:
姓名、邮箱、公司、职位/背景
带日期的互动时间线
最后接触时间戳(自动更新)
联系人评分
标签或类别
Learning system:
Maintain a learning.json config with:
skip_domains: domains to always reject (populated over time from rejections)
prefer_titles: titles that boost contact scores
skip_keywords: subject-line keywords that indicate spam
min_exchanges: minimum exchange threshold (default 1)
max_days_between: max days since last interaction (default 60)
max_attendees: meeting size cap (default 10)
min_duration_minutes: meeting length minimum (default 15)
When I reject a contact, learn from it — add their domain to skip_domains if appropriate.
学习系统:
维护一个 learning.json 配置,包含:
skip_domains:始终拒绝的域名(随着时间从拒绝中填充)
prefer_titles:提升联系人评分的职位
skip_keywords:表示垃圾邮件的主题关键词
min_exchanges:最小交流阈值(默认 1)
max_days_between:自上次互动以来的最大天数(默认 60)
max_attendees:会议规模上限(默认 10)
min_duration_minutes:会议时长最小值(默认 15)
当我拒绝一个联系人时,从中学习——如果合适,将他们的域名添加到 skip_domains。
Deduplication:
When a new contact is found, check by email, then by name+company combination.
Merge records rather than creating duplicates.
去重:
当发现新联系人时,先按邮箱检查,然后按姓名+公司组合检查。
合并记录而不是创建重复项。
Semantic retrieval:
Generate embeddings for each contact record.
Let me ask natural-language questions like:
"Who did I meet from [company] last month?"
"When did I last talk to [name]?"
"Show contacts I haven't spoken to in 30+ days."
语义检索:
为每个联系人记录生成嵌入向量。
让我可以用自然语言提问,如:
"我上个月见了 [公司] 的谁?"
"我上次和 [姓名] 交谈是什么时候?"
"显示我 30 天以上没联系过的联系人。"
Storage: SQLite with WAL mode and foreign keys enabled.
存储:启用 WAL 模式和外键的 SQLite。
Notifications: After each ingestion run, send a summary of new contacts, merges, rejections, and any issues.
通知:每次摄取运行后,发送新联系人、合并、拒绝和任何问题的摘要。
---
## 2) Knowledge Base (RAG) — Save Anything, Recall It Later
2) 知识库(RAG)——保存任何内容,稍后回忆
Build me a personal knowledge base with RAG (retrieval-augmented generation).
为我构建一个带有 RAG(检索增强生成)的个人知识库。
Ingestion — I send a URL or file and the system saves it. It should handle:
Web articles
YouTube videos (transcripts)
Tweets/X posts
PDFs
Plain text or notes
摄取——我发送一个 URL 或文件,系统保存它。它应该处理:
网页文章
YouTube 视频(转录)
推文/X 帖子
纯文本或笔记
Source type detection: Determine the type from the URL pattern or file extension. Classify as: article, video, pdf, text, tweet, or other.
源类型检测:从 URL 模式或文件扩展名确定类型。分类为:article、video、pdf、text、tweet 或 other。
Content extraction with fallback chain — this is important because no single extractor works for every site:
For Twitter/X URLs:a. Try FxTwitter API (api.fxtwitter.com) — free, no auth neededb. Fall back to X API direct lookupc. Fall back to web scraping
For YouTube URLs:a. Pull transcript via YouTube transcript API or yt-dlp
For all other URLs (articles, blogs, etc.):a. Try a clean text extractor (like Mozilla Readability or similar)b. Fall back to Firecrawl or Apify for sites that block simple extractionc. Fall back to a headless browser (Playwright/Puppeteer) for JavaScript-heavy pagesd. Last resort: raw HTTP fetch + HTML tag stripping
Retry once on transient errors (ECONNRESET, ETIMEDOUT, DNS failures) with a 2-second delay.
带降级链的内容提取——这很重要,因为没有单一提取器适用于每个网站:
对于 Twitter/X URL:a. 尝试 FxTwitter API(api.fxtwitter.com)——免费,无需认证b. 降级到 X API 直接查询c. 降级到网页抓取
对于 YouTube URL:a. 通过 YouTube 转录 API 或 yt-dlp 拉取转录
对于所有其他 URL(文章、博客等):a. 尝试干净的文本提取器(如 Mozilla Readability 或类似工具)b. 对于阻止简单提取的网站,降级到 Firecrawl 或 Apifyc. 对于重度 JavaScript 页面,降级到无头浏览器(Playwright/Puppeteer)d. 最后手段:原始 HTTP 获取 + HTML 标签剥离
对于瞬时错误(ECONNRESET、ETIMEDOUT、DNS 失败)重试一次,延迟 2 秒。
Content quality validation — reject bad extractions:
Minimum 20 characters of content.
For articles/non-tweets: at least 15% of non-empty lines must be longer than 80 characters (to detect prose vs. navigation menus).
Total content must be at least 500 characters for non-tweet types.
Detect error pages by looking for 2+ signals: "access denied", "captcha", "please enable javascript", "cloudflare", "404", "sign in", "blocked", "rate limit".
Maximum content length: 200,000 characters (truncate beyond this).
内容质量验证——拒绝糟糕的提取:
最少 20 个字符的内容。
对于文章/非推文:至少 15% 的非空行必须超过 80 个字符(以检测正文与导航菜单)。
非推文类型的总内容必须至少 500 个字符。
通过查找 2+ 个信号检测错误页面:"access denied"、"captcha"、"please enable javascript"、"cloudflare"、"404"、"sign in"、"blocked"、"rate limit"。
最大内容长度:200,000 个字符(超出部分截断)。
Deduplication — two layers:
URL-based: Normalize URLs before comparing — strip tracking params (utm_source, utm_medium, utm_campaign, fbclid, igshid, ref, s, t), remove www., normalize twitter.com to x.com, remove trailing slashes and fragments.
Content-hash: SHA-256 hash of the cleaned content. Store as a UNIQUE column — reject if the same hash already exists.
去重——两层:
基于 URL:比较前规范化 URL——去除跟踪参数(utm_source、utm_medium、utm_campaign、fbclid、igshid、ref、s、t),移除 www.,将 twitter.com 规范化为 x.com,去除尾部斜杠和片段。
内容哈希:清理后内容的 SHA-256 哈希。存储为 UNIQUE 列——如果已存在相同哈希则拒绝。
Chunking:
Chunk size: 800 characters per chunk.
Overlap: 200 characters between chunks.
Minimum chunk size: 100 characters (append tiny remainders to the last chunk).
Split on sentence boundaries (regex: /(?<=[.!?])\s+/).
分块:
块大小:每块 800 个字符。
重叠:块之间 200 个字符。
最小块大小:100 个字符(将微小余数附加到最后一块)。
按句子边界分割(正则表达式:/(?<=[.!?])\s+/)。
Embedding generation:
Use Google gemini-embedding-001 (768 dimensions, free) or OpenAI text-embedding-3-small (1536 dimensions) as fallback.
Max input: 8000 characters per chunk.
Process in batches of 10 chunks with 200ms delay between batches.
Retry failed embeddings 3 times with exponential backoff (1s, 2s, 4s).
Cache embeddings with an LRU cache (1000 entries).
嵌入生成:
使用 Google gemini-embedding-001(768 维,免费)或 OpenAI text-embedding-3-small(1536 维)作为降级方案。
最大输入:每块 8000 个字符。
以每批 10 块处理,批次之间延迟 200ms。
失败的嵌入重试 3 次,指数退避(1s、2s、4s)。
使用 LRU 缓存(1000 条目)缓存嵌入。
Storage — SQLite with two tables:
sources: id, url, title, source_type, summary, raw_content, content_hash (UNIQUE), tags (JSON array), created_at, updated_at
chunks: id, source_id (FK), chunk_index, content, embedding (BLOB), embedding_dim, embedding_provider, embedding_model, created_at
Index on chunks(source_id), sources(source_type), sources(content_hash).
Enable WAL mode and foreign keys with CASCADE deletes.
存储——带两个表的 SQLite:
sources:id、url、title、source_type、summary、raw_content、content_hash (UNIQUE)、tags (JSON array)、created_at、updated_at
chunks:id、source_id (FK)、chunk_index、content、embedding (BLOB)、embedding_dim、embedding_provider、embedding_model、created_at
在 chunks(source_id)、sources(source_type)、sources(content_hash) 上建立索引。
启用 WAL 模式和外键,带级联删除。
Concurrency protection: Use a lock file to prevent simultaneous ingestion runs. Check if lock is stale (PID dead or file older than 15 minutes).
并发保护:使用锁文件防止同时摄取运行。检查锁是否过期(PID 死亡或文件超过 15 分钟)。
Retrieval — When I ask a question:
Embed my query using the same embedding provider.
Cosine similarity search over all stored chunks. Return top 10.
Deduplicate results: keep only the best chunk per source.
Sanitize content in results (max 2500 characters per excerpt).
Pass top chunks to an LLM: "Answer using only the provided context. Cite which sources you drew from."
Return the answer with source references.
检索——当我提问时:
使用相同的嵌入提供者为我的查询生成嵌入。
对所有存储的块进行余弦相似度搜索。返回前 10 个。
结果去重:每个来源只保留最佳块。
清理结果中的内容(每段摘录最多 2500 个字符)。
将顶级块传递给 LLM:"仅使用提供的上下文回答。引用你引用的来源。"
返回答案并附带来源引用。
---
## 3) Content Idea Pipeline (Research → Dedupe → Project Management)
3) 内容创意管道(研究 → 去重 → 项目管理)
Build a content idea pipeline that researches topics, prevents duplicate ideas, and creates production-ready tasks.
构建一个内容创意管道,研究主题、防止重复创意,并创建可用于制作的任务。
Trigger: I describe a topic idea in chat.
触发:我在聊天中描述一个主题创意。
Step 1 — Research:
Search X/Twitter for recent discourse on the topic (use the tiered approach from Use Case #4 if built).
Search my knowledge base (Use Case #2) for related saved content.
Optionally run a web search.
步骤 1——研究:
搜索 X/Twitter 上关于该主题的近期讨论(如果已构建,使用用例 #4 的分层方法)。
在我的知识库(用例 #2)中搜索相关保存的内容。
可选地运行网页搜索。
Step 2 — Semantic dedupe (this is the critical gate):
Maintain a database of all past content ideas with these fields:
id (format: YYYY-MM-DD-NNN)
date, type (short/long-form), title, slug (URL-friendly, UNIQUE)
summary (the pitch text)
tags (comma-separated)
status: pitched, accepted, rejected, produced, duplicate
response (my feedback)
embedding (BLOB — generated via Gemini or OpenAI)
created_at
步骤 2——语义去重(这是关键关卡):
维护一个包含所有过去内容创意的数据库,字段如下:
id(格式:YYYY-MM-DD-NNN)
date、type(short/long-form)、title、slug(URL 友好,UNIQUE)
summary(提案文本)
tags(逗号分隔)
status:pitched、accepted、rejected、produced、duplicate
response(我的反馈)
embedding(BLOB——通过 Gemini 或 OpenAI 生成)
created_at
When a new idea comes in, run a hybrid similarity search:
Semantic similarity (70% weight): Cosine similarity between the new idea's embedding and all stored embeddings.
Keyword matching (30% weight): Match against title (30%), summary (20%), and tags (20%).
Combined score: (semantic × 0.7) + (keyword × 0.3)
当新创意进来时,运行混合相似度搜索:
语义相似度(70% 权重):新创意嵌入与所有存储嵌入之间的余弦相似度。
关键词匹配(30% 权重):与 title(30%)、summary(20%)和 tags(20%)匹配。
综合得分:(semantic × 0.7) + (keyword × 0.3)
Hard gate: If any existing idea scores above 40% combined similarity, REJECT the new idea. Show me what it matched and the similarity score. No exceptions — this prevents re-pitching old concepts.
硬门槛:如果任何现有创意综合相似度超过 40%,拒绝新创意。给我看它匹配了什么以及相似度分数。没有例外——这防止重复推销旧概念。
Step 3 — Brief assembly (only if it passes dedupe):
Keep it short — a few sentences on the angle and why it's video-worthy.
Do NOT generate titles, thumbnails, hooks, or scripts unless I explicitly ask.
Include links to relevant tweets or KB sources from the research step.
步骤 3——简报组装(仅当通过去重):
保持简短——几句关于角度和为什么值得做成视频的话。
除非我明确要求,否则不要生成标题、缩略图、钩子或脚本。
包含研究步骤中相关推文或知识库来源的链接。
Step 4 — Create task:
Create a task in my project management tool (Asana, Linear, Notion, or Todoist) with the brief as the description.
步骤 4——创建任务:
在我的项目管理工具(Asana、Linear、Notion 或 Todoist)中创建一个任务,简报作为描述。
Step 5 — Store the idea:
Add the new idea + its embedding to the pitch database so future ideas are compared against it.
Set initial status to "pitched".
步骤 5——存储创意:
将新创意 + 其嵌入添加到提案数据库,以便将来的创意与之比较。
设置初始状态为 "pitched"。
Step 6 — Notify me with the brief and a link to the created task.
步骤 6——用简报和创建任务的链接通知我。
---
## 4) Social Media Research System (Cost-Optimized)
4) 社交媒体研究系统(成本优化)
Build a social media research tool that answers "What are people saying about [topic]?" using X/Twitter, while minimizing API costs.
构建一个社交媒体研究工具,使用 X/Twitter 回答"人们在说什么关于 [主题]?",同时最小化 API 成本。
Query decomposition:
Take my question and break it into 2-4 focused search queries covering different angles.
查询分解:
获取我的问题并将其分解为 2-4 个涵盖不同角度的聚焦搜索查询。
Tiered retrieval — always try the cheapest option first:
Tier 1 (free): FxTwitter API (api.fxtwitter.com) — works for individual tweet lookups, no auth needed.
Tier 2 (low-cost, ~$0.15/1K tweets): A paid Twitter data provider like TwitterAPI.io or SocialData. Supports search, profiles, user tweets, and thread context.
Tier 3 (expensive, ~$0.005/tweet): Official X API v2 (api.x.com/2). Last resort only. Rate limit to 350ms between requests to stay under 450 req/15min.
分层检索——始终先尝试最便宜的选项:
第一层(免费):FxTwitter API(api.fxtwitter.com)——适用于单个推文查询,无需认证。
第二层(低成本,约 $0.15/1K 推文):付费 Twitter 数据提供商如 TwitterAPI.io 或 SocialData。支持搜索、个人资料、用户推文和线程上下文。
第三层(昂贵,约 $0.005/推文):官方 X API v2(api.x.com/2)。仅作最后手段。请求之间限速 350ms 以保持在 450 请求/15 分钟以下。
How the tiers cascade by operation:
Single tweet lookup: Tier 1 → Tier 2 → Tier 3
Search: Tier 2 → Tier 3
Profile lookup: Tier 2 → Tier 3
Thread expansion: Tier 2 → Tier 3
各操作的分层降级方式:
单个推文查询:第一层 → 第二层 → 第三层
搜索:第二层 → 第三层
个人资料查询:第二层 → 第三层
线程扩展:第二层 → 第三层
Filtering:
Filter by timeframe (default: last 7 days, configurable).
Rank by engagement (likes + retweets + replies).
Remove retweets and duplicate content.
Suppress low-quality or spam results.
过滤:
按时间范围过滤(默认:过去 7 天,可配置)。
按互动量排序(点赞 + 转发 + 回复)。
移除转发和重复内容。
抑制低质量或垃圾结果。
Thread expansion: For high-engagement tweets that are part of a thread, pull the full thread.
线程扩展:对于属于线程的高互动推文,拉取完整线程。
Caching: Cache results with a 1-hour TTL so repeated queries don't burn API credits.
缓存:缓存结果,TTL 为 1 小时,以便重复查询不会消耗 API 额度。
Usage logging: Log every API call to a separate log file per tier, with timestamps and estimated costs.
使用日志:将每个 API 调用记录到每层单独的日志文件中,带时间戳和预估成本。
Output — synthesize into a briefing:
Key narratives: 3-5 dominant takes
Notable posts: 5-10 most relevant tweets with links
Sentiment summary: positive, negative, or mixed
Contrarian takes: interesting minority opinions
输出——综合成简报:
关键叙事:3-5 个主导观点
值得注意的帖子:5-10 条最相关的推文及链接
情感摘要:正面、负面或混合
逆向观点:有趣的少数派意见
---
## 5) YouTube Analytics + Competitor Tracking
5) YouTube 分析 + 竞争对手跟踪
Build a YouTube analytics system that tracks my channel daily and monitors competitors.
构建一个 YouTube 分析系统,每天跟踪我的频道并监控竞争对手。
My channel — collect these metrics daily via YouTube Data API and YouTube Analytics API:
views (total and per-video)
estimatedMinutesWatched (watch time)
averageViewDuration
subscribersGained
impressions and CTR (if available via Analytics API)
我的频道——每天通过 YouTube Data API 和 YouTube Analytics API 收集这些指标:
views(总计和每个视频)
estimatedMinutesWatched(观看时长)
averageViewDuration
subscribersGained
impressions 和 CTR(如果 Analytics API 可用)
Store each day's metrics in a SQLite table:
daily_stats: video_id, date, views, watch_time_minutes, avg_view_duration, impressions, ctr, subscribers_gained
videos: id, title, published_at, thumbnail_url, duration_seconds
将每天的指标存储在 SQLite 表中:
daily_stats:video_id、date、views、watch_time_minutes、avg_view_duration、impressions、ctr、subscribers_gained
videos:id、title、published_at、thumbnail_url、duration_seconds
Derived metrics:
7-day moving average for daily views
Subscriber-to-view conversion rate
Views-per-video trend
派生指标:
每日浏览量的 7 天移动平均
订阅者到浏览量的转化率
每个视频浏览量趋势
Competitor tracking:
Accept a configurable list of competitor channel IDs/usernames.
Daily, pull each competitor's recent uploads (title, publish date, views) and subscriber count via YouTube Data API.
Track upload cadence and momentum.
竞争对手跟踪:
接受可配置的竞争对手频道 ID/用户名列表。
每天通过 YouTube Data API 拉取每个竞争对手的近期上传(标题、发布日期、浏览量)和订阅者数量。
跟踪上传节奏和势头。
Chart generation — produce PNG charts using matplotlib (dark theme):
"trend" chart: Daily view trend for a specific video (line chart with fill area)
"top" chart: Top N videos by views (horizontal bar chart, color-coded by short vs long-form)
"daily" chart: Channel-wide daily views with 7-day moving average overlay
Format numbers with commas, dates as readable labels, include grid and legends.
Save charts to a designated output folder.
图表生成——使用 matplotlib 生成 PNG 图表(深色主题):
"trend" 图表:特定视频的每日浏览量趋势(带填充区域的折线图)
"top" 图表:按浏览量排名的前 N 个视频(水平条形图,按短视频与长视频着色)
"daily" 图表:频道范围的每日浏览量,带 7 天移动平均叠加
数字用逗号格式化,日期作为可读标签,包含网格和图例。
将图表保存到指定输出文件夹。
Schedule: Daily cron job collecting all data. Optional daily summary sent to chat.
计划:每天定时任务收集所有数据。可选地发送每日摘要到聊天。
---
## 6) Nightly Business Briefing (Multi-Perspective AI Council)
6) 夜间业务简报(多视角 AI 委员会)
Build a nightly business analysis system that collects signals from across my tools and produces prioritized recommendations using a multi-persona AI review council.
构建一个夜间业务分析系统,从各种工具收集信号,并使用多角色 AI 审查委员会生成优先推荐。
Signal collection — pull data from whatever systems you use. Normalize each signal into:
{ source, signal_name, value, confidence (0-100), direction (up/down/flat), category }
信号收集——从你使用的任何系统拉取数据。将每个信号规范化为:
{ source, signal_name, value, confidence (0-100), direction (up/down/flat), category }
Example sources: YouTube metrics, CRM health, project management backlog, social growth, email themes, meeting action items, sales pipeline, operations/cron reliability.
示例来源:YouTube 指标、CRM 健康度、项目管理待办事项、社交增长、邮件主题、会议行动项、销售管道、运营/定时任务可靠性。
Compact to the top 200 signals by confidence score, with up to 4 sample events per source.
按置信度分数压缩到前 200 个信号,每个来源最多 4 个样本事件。
Three-phase AI review council (all phases use a frontier model like Claude Opus or GPT-4):
三阶段 AI 审查委员会(所有阶段使用前沿模型如 Claude Opus 或 GPT-4):
Phase 1 — Draft (LeadAnalyst):
Score business outlook on weekly and monthly horizons (0-100), plus a blended average.
Generate 5-10 recommendations, each with: title, description (2-3 plain sentences), evidence references, estimated impact (0-100), estimated effort (0-100), confidence (0-100).
Constraints: use only provided data, no made-up facts, plain human language, don't recommend publishing content immediately.
阶段 1——起草(LeadAnalyst):
对每周和每月的业务前景进行评分(0-100),外加混合平均值。
生成 5-10 条推荐,每条包含:标题、描述(2-3 句简单句子)、证据引用、预估影响(0-100)、预估工作量(0-100)、置信度(0-100)。
约束:仅使用提供的数据,不编造事实,简单人类语言,不推荐立即发布内容。
Phase 2 — Parallel review (4 personas, run simultaneously via Promise.all):
GrowthStrategist: Scalable growth, asymmetric upside. Demands concrete evidence.
RevenueGuardian: Near-term revenue and cash flow. Challenges anything that doesn't protect income.
SkepticalOperator: Execution reality, data quality risks. Challenges weak assumptions.
TeamDynamicsArchitect: Team health, collaboration quality, sustainable pace.
阶段 2——并行审查(4 个角色,通过 Promise.all 同时运行):
GrowthStrategist:可扩展增长,非对称上行空间。要求具体证据。
RevenueGuardian:近期收入和现金流。挑战任何不保护收入的事情。
SkepticalOperator:执行现实、数据质量风险。挑战薄弱假设。
TeamDynamicsArchitect:团队健康、协作质量、可持续节奏。
Each reviewer:
Critiques every recommendation: support, revise, or reject (with score adjustments)
Can propose new recommendations the draft missed
Returns: reviewer_name, key_findings[], votes[] (with adjusted scores), new_recommendations[]
每个审查者:
审查每条推荐:支持、修订或拒绝(带分数调整)
可以提出起草遗漏的新推荐
返回:reviewer_name、key_findings[]、votes[](带调整后的分数)、new_recommendations[]
Phase 3 — Consensus (CouncilModerator):
Receives draft + all 4 reviews.
Reconciles disagreements.
Produces final recommendation set with consensus notes.
阶段 3——共识(CouncilModerator):
接收起草 + 所有 4 份审查。
协调分歧。
生成带共识说明的最终推荐集。
Fallback: If consensus fails, use draft recommendations. If a reviewer fails, substitute a stub review with the error message.
降级:如果共识失败,使用起草推荐。如果审查者失败,用错误消息替代存根审查。
Ranking — score each recommendation:
priority = (impact × w1) + (confidence × w2) + ((100 - effort) × w3)
Default weights: w1=0.4, w2=0.35, w3=0.25. Store weights in a policy table and update via feedback over time.
排序——为每条推荐评分:
priority = (impact × w1) + (confidence × w2) + ((100 - effort) × w3)
默认权重:w1=0.4、w2=0.35、w3=0.25。将权重存储在策略表中,并通过反馈随时间更新。
Filter: Remove any "publish now" recommendations (hard constraint).
过滤:移除任何"立即发布"的推荐(硬约束)。
Output: Deliver ranked recommendations to my preferred channel. Store full council trace (draft, all reviews, consensus, models used) in a database for auditing and backtesting.
输出:将排序后的推荐发送到我的首选渠道。将完整的委员会追踪(起草、所有审查、共识、使用的模型)存储在数据库中用于审计和回测。
Schedule: Run nightly via cron.
计划:通过定时任务每晚运行。
---
7) CRM/Business Tool Natural Language Access
CRM/业务工具自然语言访问
Build a natural-language interface to my CRM (HubSpot, Salesforce, or whatever I use).
为我的 CRM(HubSpot、Salesforce 或我使用的任何工具)构建自然语言界面。
Supported objects and operations:
- Contacts: search, create, update, list, get by ID
- Companies: search, create, update, list
- Deals: search, create, update, list, query by pipeline stage
- Owners: list
- Associations: link contacts to companies, deals to contacts, deals to companies
(For HubSpot, common association type IDs: 1=Contact→Company, 3=Deal→Contact, 5=Deal→Company)
- Properties/schema: inspect available fields for any object
支持的对象和操作:
- Contacts:search、create、update、list、get by ID
- Companies:search、create、update、list
- Deals:search、create、update、list、query by pipeline stage
- Owners:list
- Associations:将 contacts 链接到 companies、deals 到 contacts、deals 到 companies
(对于 HubSpot,常见的关联类型 ID:1=Contact→Company、3=Deal→Contact、5=Deal→Company)
- Properties/schema:检查任何对象的可用字段
Intent classification — parse my message into:
1. Lookup: "Find the contact for [name]" → search contacts
2. Create: "Add a new contact: [name], [email], [company]" → create contact
3. Update: "Update [contact]'s title to [new title]" → update contact
4. List: "Show all deals in negotiation stage" → filter deals
5. Associate: "Link [contact] to [company]" → create association
意图分类——将我的消息解析为:
1. Lookup:"Find the contact for [name]" → search contacts
2. Create:"Add a new contact: [name], [email], [company]" → create contact
3. Update:"Update [contact]'s title to [new title]" → update contact
4. List:"Show all deals in negotiation stage" → filter deals
5. Associate:"Link [contact] to [company]" → create association
Validation: If I'm missing required fields, ask me before proceeding.
验证:如果我缺少必填字段,在继续之前询问我。
Response format: Return clean, readable summaries — not raw JSON. For deal queries, include: deal name, amount, stage, close date, last modified.
响应格式:返回干净、可读的摘要——不是原始 JSON。对于 deal 查询,包含:deal name、amount、stage、close date、last modified。
Authentication: API key or OAuth stored in environment variables.
认证:API 密钥或 OAuth 存储在环境变量中。
8) AI Content Humanization
AI 内容人性化
Build a text rewriting tool that removes AI-generated artifacts and makes content sound like a real person wrote it.
构建一个文本重写工具,移除 AI 生成的痕迹,使内容听起来像真人写的。
Input: I paste draft text (social posts, blog paragraphs, emails, video scripts).
输入:我粘贴草稿文本(社交帖子、博客段落、邮件、视频脚本)。
Step 1 — Detection. Scan for common AI tells:
- Overuse of: "delve", "landscape", "leverage", "it's important to note", "in conclusion", "game-changing", "revolutionary", "transformative"
- Tone inflation: dramatic language when the subject doesn't warrant it
- Generic phrasing that could apply to any topic
- Repetitive sentence structures (every sentence starting the same way)
- Excessive hedging: "It's worth noting that perhaps..."
- Lists that feel generated (too clean, too parallel, no personality)
- Identical paragraph lengths and rhythms
步骤 1——检测。扫描常见的 AI 痕迹:
- 过度使用:"delve"、"landscape"、"leverage"、"it's important to note"、"in conclusion"、"game-changing"、"revolutionary"、"transformative"
- 语调膨胀:当主题不值得时使用夸张语言
- 可适用于任何主题的通用措辞
- 重复的句子结构(每个句子以相同方式开头)
- 过度委婉:"It's worth noting that perhaps..."
- 感觉生成的列表(太干净、太平行、没有个性)
- 相同的段落长度和节奏
Step 2 — Rewrite:
- Replace vague qualifiers with specific, concrete language
- Vary sentence length (mix short punchy sentences with longer ones)
- Use contractions, sentence fragments, and informal word choices where natural
- Remove filler while keeping the core message
- Add the kind of imperfections real writing has — not errors, but human cadence
步骤 2——重写:
- 用具体、实在的语言替换模糊的限定词
- 变化句子长度(混合短促有力的句子和较长的句子)
- 在自然的地方使用缩略语、句子片段和非正式用词
- 移除填充内容同时保留核心信息
- 添加真实写作所具有的那种不完美——不是错误,而是人类的节奏
Step 3 — Channel tuning (optional, based on destination):
- Twitter/X: Punchy, under 280 chars, direct, no filler
LinkedIn: Professional but conversational, authoritative without being stiff
Blog: Longer form, personal anecdotes and opinions welcome
Email: Brief, clear, action-oriented, warm but efficient
步骤 3——渠道调优(可选,基于目标):
Twitter/X:有力、少于 280 字符、直接、无填充
LinkedIn:专业但对话式,权威但不生硬
博客:较长形式,欢迎个人轶事和观点
邮件:简短、清晰、以行动为导向、温暖但高效
Output: Return the revised text. Optionally highlight what changed and why.
输出:返回修订后的文本。可选地高亮显示更改了什么以及为什么。
---
## 9) Image Generation + Iterative Editing
9) 图像生成 + 迭代编辑
Build an image generation workflow where I describe what I need in chat and iterate until it's right.
构建一个图像生成工作流,我在聊天中描述我需要什么并迭代直到满意。
Core flow:
I describe an image: subject, style, composition, mood, colors, any text to include.
Generate 1-3 variants using an image generation API (DALL-E 3, Midjourney API, Stable Diffusion, or Flux).
I review and either:a. Accept → save as final assetb. Request changes → I describe adjustments ("darker background", "remove text", "more minimal") and it generates new variants incorporating my feedback
Loop continues until I'm satisfied.
核心流程:
我描述图像:主题、风格、构图、氛围、颜色、要包含的任何文本。
使用图像生成 API(DALL-E 3、Midjourney API、Stable Diffusion 或 Flux)生成 1-3 个变体。
我审查并选择:a. 接受 → 保存为最终资源b. 请求更改 → 我描述调整("darker background"、"remove text"、"more minimal"),它生成包含我反馈的新变体
循环继续直到我满意。
Image editing:
I can send an existing image and ask for edits: swap background, adjust composition, combine multiple images.
Use inpainting or img2img when available.
图像编辑:
我可以发送现有图像并要求编辑:更换背景、调整构图、组合多个图像。
可用时使用修复或 img2img。
Context tracking: Remember what we've been working on across multiple messages in the same session so I don't have to re-describe the whole concept each iteration.
上下文跟踪:记住在同一会话中多条消息中我们一直在处理的内容,这样我就不必在每次迭代时重新描述整个概念。
Output: Save final assets to a designated folder. Deliver as download
输出:将最终资源保存到指定文件夹。作为下载提供