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 show-today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。
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
正在显示 SKILL.md
| name | show-today |
| description | Priority tasks for today — checks live sources (email, calendar) before showing stale CSV data |
What to do today — reality-checked against live sources
email-read, timezone| What | Path |
|---|---|
| Tasks | $DATA_PATH/projects/pm/pm_tasks_master.csv |
| Projects | $DATA_PATH/projects/pm/pm_projects_master.csv |
| Leads | $DATA_PATH/sales/crm/relationships/leads.csv |
| People | $DATA_PATH/sales/crm/contacts/people.csv |
| Calendar script | $GOOGLE_TOOLS_PATH/read_calendar.py |
| Email script | $GOOGLE_TOOLS_PATH/read_emails.py |
| Python venv | $GOOGLE_TOOLS_PATH/.venv/bin/python3 |
CRITICAL: Always gather live state BEFORE reading CSVs.
Step 1: Live sources → calendar + inbox + sent (last 3 days)
Step 2: Static sources → leads.csv + pm_tasks_master.csv
Step 3: Reconcile → cross-reference emails with lead contacts
Step 4: Display → show corrected picture, grouped by priority tier
Run all three in parallel:
# Calendar — today's events
$GOOGLE_TOOLS_PATH/.venv/bin/python3 $GOOGLE_TOOLS_PATH/read_calendar.py 1
# Inbox — last 3 days, all messages
$GOOGLE_TOOLS_PATH/.venv/bin/python3 $GOOGLE_TOOLS_PATH/read_emails.py 30 "after:$(date -v-3d '+%Y/%m/%d')"
# Sent — last 3 days
$GOOGLE_TOOLS_PATH/.venv/bin/python3 $GOOGLE_TOOLS_PATH/read_emails.py 20 "in:sent after:$(date -v-3d '+%Y/%m/%d')"
import pandas as pd
from datetime import date
today = str(date.today())
CRM = '$DATA_PATH/sales/crm'
PM = '$DATA_PATH/projects/pm'
leads = pd.read_csv(f'{CRM}/relationships/leads.csv')
people = pd.read_csv(f'{CRM}/contacts/people.csv')
tasks = pd.read_csv(f'{PM}/pm_tasks_master.csv')
# Active leads only (not won/lost/dead)
active_leads = leads[leads['stage'].isin(['new', 'qualified', 'proposal', 'negotiation'])]
# Leads with actions due today or overdue
due = active_leads[active_leads['next_action_date'] <= today]
For each active lead that has next_action containing "waiting"/"check"/"follow-up":
people.csvImplementation:
# Build email lookup: person_id -> email
contact_emails = dict(zip(people['person_id'], people['email']))
# For each due lead, check if we have a reply
for _, lead in due.iterrows():
contact_id = lead['primary_contact_id']
email = contact_emails.get(contact_id, '')
if not email or pd.isna(email):
continue
# Check if this email appears in recent inbox
# (Claude: search the inbox output from Step 1 for this email address)
# If found: flag as STALE — reply exists, next_action is outdated
# If not found: lead state is current
Claude does this manually — scan Step 1 inbox output for email addresses matching lead contacts.
IMPORTANT: Use priority tiers, not raw priority_score.
=== TODAY: {date} ===
CALENDAR:
{list today's events with times in WITA}
---
STALE ALERTS (leads where reply already exists but CRM not updated):
- [lead-id] {contact name} replied on {date} — CRM says "{stale next_action}"
ACTION NEEDED: read email, update lead, respond
---
MONEY NOW (invoices, payments, delivery deadlines):
- {items}
MONEY SOON (deals in negotiation/proposal with active momentum):
- {items}
LEADS — ACTION TODAY:
- {leads with next_action_date == today}
LEADS — OVERDUE (skip anything older than 14 days unless high priority):
- {leads with next_action_date < today, max 14 days old}
PM TASKS (in_progress or hot):
- {items, skip auto-generated email reply tasks}
pm_tasks_master.csv entries with task_name starting with "Reply:" are auto-generated by email agent. Do NOT show them as today's tasks.next_action_date is >14 days ago and no recent email activity, skip. Mention count: "N stale leads hidden (run query-leads to review)."CRM next_action contains | Check inbox for | If found → |
|---|---|---|
| "waiting for reply" / "check reply" | emails FROM contact | Reply exists — read and respond |
| "follow-up" / "ping" / "nudge" | emails FROM contact | They replied before we pinged — update action |
| "send proposal" / "send NDA" | emails in SENT to contact | Already sent — update to "waiting for response" |
| "schedule call" / "book meeting" | calendar events with contact name | Already booked — update to call prep |
When showing tasks and user says "done" or it's clear a task was completed:
pm-done skill)=== TODAY: 2026-03-03 ===
CALENDAR:
16:00 Client H x WelabelData (AI testing)
18:00 Alex/Ivan (discovery call)
21:30 Discovery Call: Frank Castle (Medical AI)
---
STALE ALERTS:
- [lead-clienti-001] Edward Norton replied 2x (01:22, 02:47) —
CRM says "Чекаємо відповідь, дьорнути якщо тиша до 05.03"
ACTION: Edward asking for formal proposal on letterhead. Respond today.
---
MONEY NOW:
- Invoice #99 Enterprise Corp/Client C ($X,XXX) — 19 days overdue, Alice checking fee structure
- Mercury account closure — withdraw funds by April 16
MONEY SOON:
- Client I — Edward wants formal proposal (letterhead, payment terms, dispute resolution)
- Client G NDA — print, sign, send scan to Grace today
LEADS — ACTION TODAY:
- [lead-example-001] Discovery call 18:00 WITA
- [lead-clientg-001] Sign and send NDA
(3 stale leads hidden — run query-leads to review)
email-read — read inbox/sent for live statepm-done — mark task done + create follow-upquery-leads — detailed lead queries, bulk review stale leadsweekly-review — weekly progress reporttimezone — time conversion for calendar events