用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/aAAaqwq/AGI-Super-Team --skill dispatcher命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
币安广场合约投机雷达 v5:以最近24小时专业交易帖为主要证据,回源核验帖子, 联合币安公共合约行情、4周期K线、布林带、ATR、量能和RR,生成可审计的本地影子报告。 触发词:币安广场、扫描币安、binance square、合约机会、交易信号雷达、4小时雷达
BTC 5分钟K线实时方向预测。v5.9对抗式审查重构: 13因子收敛到3个有证据信号(half_body延续+volume放量+meanrev回归, 11个47-49%硬币因子清零) + 三层独立信息过滤(多周期4h/1h/15m趋势 + 跨资产ETH/SOL广度 + 真订单流OFI) + 移除bull×0.92惩罚/Platt置信度门控。半K线策略第2分钟执行。黑天鹅防护: ATR spike+FNG<25。Binance端点双向故障切换。
BB 双向套利策略:加密合约 10x 杠杆布林带均值回归。布林带收窄=横盘→在下轨买、上轨卖;三重过滤器(1h趋势/RSI/BB甜区)确认碗平放,轨对轨止盈(RR 2:1~4:1)。含实时WebSocket模拟盘(paper)、历史回测(simulate/backtest_daily)、币安永续实盘CLI(trade_exec)。触发:'bb套利'、'布林带'、'bollinger'、'横盘策略'、'NEAR'、'回测'、'模拟盘'、'paper trading'。
基于 SOC 职业分类
| name | dispatcher |
| description | Route complex multi-step tasks, multi-skill orchestration, post-execution checklist |
| user-invocable | false |
MANDATORY first step before selecting any skill. Contains the complete routing table for all 61+ skills. The system prompt only shows ~50 skills — this file is the single source of truth.
Claude Code has a ~50 skill limit in the system prompt. Skills beyond this limit are silently dropped. The dispatcher contains the COMPLETE routing table. Always consult this file before choosing a skill, especially for:
email-send-direct (NOT email-send-bulk)call-prepdocumentsDetermine:
| Trigger | Skill |
|---|---|
| Add new company/person/lead | add-lead |
| Update lead status, notes, contacts | update-lead |
| Search, filter, reports on CRM data | query-leads |
| Import data from staging to CRM | crm-import |
| Log email/call/meeting activity | log-activity |
| Validate CRM/PM changes before PR | change-review |
| Trigger | Skill |
|---|---|
| "What should I do today?" / priority tasks | show-today |
| Morning briefing (tasks + email + follow-ups) | daily-briefing |
| Weekly project review report | weekly-review |
| Create new project with task breakdown | create-project |
| Mark task done + schedule follow-up | pm-done |
| Task prioritization and scoring | task-prioritization |
| Trigger | Skill |
|---|---|
| Prepare for a call with client/lead | call-prep |
| Analyze client requests → scoping + pricing | client-discovery |
| Create client folder, share docs, pre-call check | client-workspace |
| Multi-channel outreach (TG + email + WhatsApp) | mass-outreach |
| Trigger | Skill |
|---|---|
| Send single email (with dry-run, reply, attachments) | email-send-direct |
| Bulk email sending | email-send-bulk |
| Read inbox / sent emails via Gmail API | email-read |
| AI inbox classification + TG notification | email-monitor |
| Send Telegram DM (CSV, rate-limited) | telegram-send |
| Check inbound Telegram messages | telegram-check |
| Export/import/lookup Telegram contacts | telegram-contacts |
| Telegram group management (post, members) | telegram-groups |
| Search Telegram channels, read posts | telegram-scrape |
| Send WhatsApp message (Baileys) | whatsapp-send |
| LinkedIn automation (mouse/keyboard, CDP) | linkedin-cdp |
| Connect to AI agent via MCP | mcp-agent-connect |
| Trigger | Skill |
|---|---|
| Generate invoice (manual, no CRM) | invoice |
| Auto-generate invoice with CRM integration | invoice-generator-agent |
| Track expenses, export data | expense-tracker |
| Stripe webhook management (AI Kitchen Pro) | stripe-webhook |
| Trigger | Skill |
|---|---|
| AI code review for PR or local changes | code-review |
| Git: branch, PR, merge, cleanup | git-workflow |
| Deploy website via GitHub PR + Cloud Build | deploy-website |
| Trigger | Skill |
|---|---|
| Google OAuth setup / refresh tokens | google-auth |
| Google Drive: upload, folders, share, create doc | google-drive |
| Facebook CDP automation (comments, data) | infra-facebook-cdp |
| Create/update Telethon session | telegram-session |
| ttyd + Tailscale for mobile terminal access | remote-access |
| Trigger | Skill |
|---|---|
| Review NDA, contracts, agreements | legal-review |
| Trigger | Skill |
|---|---|
| Build agent from spec | agent-builder |
| Test agent (dry-run, unit, integration) | agent-tester |
| Process analysis, gap finding, spec generation | process-analyst |
| Auto: daily briefing | daily-briefing-run |
| Auto: weekly review | weekly-review-run |
| Auto: email outreach | email-outreach-run |
| Auto: Telegram inbound processing | telegram-inbound-run |
| Auto: Telegram scraping | telegram-scraper-run |
| Auto: LinkedIn inbound processing | linkedin-inbound-run |
| Auto: WhatsApp outreach | whatsapp-outreach-run |
| Auto: channel/contact sync | channel-truth-run |
| Auto: lead trigger monitoring | watchers-run |
| Auto: payment tracking | payment-tracker-run |
| Auto: experiment runner | experiment-runner-run |
| Trigger | Skill |
|---|---|
| Send rekvizity, passport data, INN | documents |
| Upload document scans to Drive | documents |
| "де мої документи?" / company data | documents |
| Trigger | Skill |
|---|---|
| Persistent memory (observations, summaries) | memory |
| Timezone conversion, local time | timezone |
| Company knowledge base (domains, infra, accounts) | company-wiki |
| AI agent contacts (add, list, remove MCP) | agent-contacts |
Examples of combinations:
"Send all clients a message about new prices"
1. query-leads → get clients (relationships/clients.csv)
2. For each → find contact (contacts/people.csv)
3. telegram-send / email-send-bulk → send
4. log-activity → record in activities.csv
"Good morning, what's new?"
1. daily-briefing → tasks + email + follow-ups
"Prepare for a call with X and send the agenda"
1. call-prep → research + plan (includes workspace check)
2. email-send-direct → send agenda
3. log-activity → record activity
"Client sent a list of automation requests"
1. query-leads → get CRM context
2. client-discovery → analyze requests, estimate hours, prioritize
3. client-workspace → create shared questionnaire doc
4. log-activity → record activity
"Set up folder for new client"
1. query-leads → get company + contact email
2. client-workspace → create folder structure + share
3. log-activity → record activity
"Generate and send invoice to client"
1. query-leads → get client + billing info
2. invoice-generator-agent → generate invoice
3. google-drive → upload to client folder
4. email-send-direct → send to client
5. log-activity → record
"Deploy website changes"
1. code-review → check changes
2. git-workflow → branch + PR
3. deploy-website → deploy via Cloud Build
"Review this NDA and send feedback"
1. legal-review → analyze document
2. email-send-direct → send feedback
3. log-activity → record
"Send rekvizity / passport data to partner"
1. documents → read structured data blocks
2. email-send-direct → send (reply to thread)
3. log-activity → record
"User provides document scans"
1. documents → extract data, upload scans to Drive
2. Update passport_data.md with structured data
After executing a task that modifies data:
What changed?
│
├── DATA (CSV in CRM/PM)?
│ ↓
│ change-review → git-workflow
│
└── CODE (*.py, *.js, *.md)?
↓
code-review → git-workflow
If there was outreach (email/telegram/call):
activities.csv (log-activity)pm_tasks_master.csvIf CRM/PM data was modified:
schema.yamlIf code or skills were modified:
| Domain | Path |
|---|---|
| CRM | $CRM_PATH/ |
| PM | $PM_PATH/ |
| CRM Schema | sales/crm/schema.yaml (always read first for CRM tasks) |
When creating a new skill, ALWAYS complete ALL steps:
Create SKILL.md in ~/claude-skills/skills/{category}/{skill-name}/SKILL.md
~/claude-skills/skills/TEMPLATE.mdname and descriptionCreate symlink: ln -s ~/claude-skills/skills/{category}/{skill-name} ~/.claude/skills/{skill-name}
Update THIS dispatcher:
Verify discovery:
Commit via git-workflow: branch → PR → merge
CRITICAL: Steps 3-4 are what gets forgotten. If you skip updating the dispatcher, the skill becomes invisible when the system prompt drops it.