docs: scenarios as agent prompts, not CLI commands

The user talks to their agent in natural language. The agent uses the
skill internally. Scenarios should show what you SAY to your agent,
not what wxclawbot commands to run. The CLI reference section is for
the agent and scripts, not for humans.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: lroolle

README.en.md

@@ -2,81 +2,61 @@
 
 [简体中文](./README.md) | [npm](https://www.npmjs.com/package/@herai/wxclawbot-cli) | [GitHub](https://github.com/lroolle/wxclawbot-cli) | [ClawHub](https://clawhub.ai/lroolle/wxclawbot-send)
 
-Let your OpenClaw AI agent proactively send WeChat messages. Text, images, video, files -- whatever you need.
+Let your OpenClaw AI agent proactively send WeChat messages. Text, images, video, files.
 
 ## Why This Exists
 
-WeChat's official bot API has one glaring limitation: bots can only *reply*. They can't initiate conversations. It's like hiring an assistant who only speaks when spoken to. Completely useless for reminders, alerts, or anything proactive.
+WeChat's bot API only supports replies -- bots can't initiate. This skill fixes that. Your agent can now send messages on its own schedule.
 
 > "Currently doesn't support proactively sending you messages on a schedule"
 >
 > -- [Lobster's WeChat Integration Tutorial](https://mp.weixin.qq.com/s/nYDQ1obQEHe1WavGpNzasQ)
 
-**This CLI fixes that.** Your OpenClaw agent can now send messages on its own schedule -- timed reminders, monitoring alerts, automated reports. Reuses [openclaw-weixin](https://www.npmjs.com/package/@tencent-weixin/openclaw-weixin) credentials, no extra auth dance.
+## How It Works
 
-## What Can It Actually Do?
-
-### Life Reminders (Before Your Body Reminds You the Hard Way)
+### 1. Install the skill
 
 ```bash
-# Every 45 minutes
-wxclawbot send --text "Drink water. Don't wait for kidney stones to remind you." --json
-
-# Every hour of sitting
-wxclawbot send --text "Your ass has been welded to that chair for 60 minutes. Stand up." --json
-
-# 1 AM and still coding
-wxclawbot send --text "Your code quality is receding faster than your hairline. Go to sleep." --json
-
-# Friday 5:55 PM
-wxclawbot send --text "5 minutes to freedom. Don't accept new tasks. Don't reply to messages. Prepare to run." --json
-
-# Weekdays 11:15 AM
-wxclawbot send --text "Order lunch now or delivery gets backed up until 2 PM. Enjoy hunger-driven development." --json
+clawhub install wxclawbot-send
 ```
 
-### DevOps (Know When Production Is on Fire)
-
-```bash
-# CI/CD broke
-wxclawbot send --text "CI is down. It was @zhangsan's commit abc1234. Please clean up your mess." --json
-
-# PR sitting unreviewed for 24h
-wxclawbot send --text "PR #42 has been sitting there for 24 hours. Review it. Code doesn't merge itself." --json
+### 2. Just talk to your agent
 
-# Deploy succeeded
-wxclawbot send --text "v2.3.1 is live. Nothing has exploded yet." --json
+No code, no cron config, no CLI. The agent handles scheduling and invocation internally.
 
-# Deploy failed
-wxclawbot send --text "v2.3.1 blew up. Don't panic. Well, panicking won't help anyway." --json
+**Life reminders:**
 
-# Error rate spike
-wxclawbot send --text "Production error rate at 15%. Normal is 0.1%. Might want to look at that." --json
-
-# GitHub contributions gap
-wxclawbot send --text "Your GitHub contribution graph is turning into a desert. 3 days without commits." --json
-```
-
-### Business Operations (Let Robots Do the Boring Stuff)
-
-```bash
-# Daily report
-wxclawbot send --file ./daily-report.pdf --text "Today's numbers. Attached." --json
+> Remind me to drink water every 45 minutes, be blunt about it
+>
+> Every hour of sitting, tell me to stand up
+>
+> If I'm still chatting with you after 1 AM, tell me to go to sleep
+>
+> Friday 5:55 PM, remind me to stop working and leave
+>
+> Weekdays 11:15, remind me to order lunch before delivery backs up
 
-# SLA breach
-wxclawbot send --text "Ticket #1024 has exceeded SLA by 4 hours. Escalating." --json
+**DevOps alerts:**
 
-# Security alert
-wxclawbot send --text "Suspicious login detected from Ho Chi Minh City. If that's not you, change your password now." --json
+> When CI breaks, WeChat me immediately with whose commit caused it
+>
+> If a PR sits unreviewed for 24 hours, ping me
+>
+> Notify me on every deploy, success or failure
+>
+> If production error rate exceeds 1%, alert me via WeChat immediately
 
-# Server alert
-wxclawbot send --text "prod-03 disk usage at 95%. Clean it up or enjoy the disk-full show." --json
+**Business ops:**
 
-# Screenshots / images
-wxclawbot send --file ./monitoring-screenshot.png --text "Dashboard snapshot" --json
-```
+> Send me yesterday's key metrics summary every morning at 9
+>
+> Auto-escalate tickets that exceed SLA by 4 hours
+>
+> Alert me on suspicious logins immediately
+>
+> Notify me when any server disk exceeds 90%
 
-Combine with cron, GitHub Actions, or your own monitoring scripts. Go wild.
+That's it. You talk, the agent works.
 
 ## Install
 
@@ -90,9 +70,9 @@ Or via npm:
 npm install -g @herai/wxclawbot-cli
 ```
 
-Requires Node.js >= 20. This is not negotiable.
+Requires Node.js >= 20, openclaw-weixin logged in.
 
-## Usage
+## CLI Reference (for agents and scripts)
 
 ```bash
 wxclawbot send --text "message" --json
@@ -102,45 +82,31 @@ wxclawbot send --to "user@im.wechat" --text "Hello" --json
 echo "report ready" | wxclawbot send --json
 ```
 
-- `--to` defaults to the bound user from your openclaw account. Use `--to` for someone else.
-- `--json` -- **always use this**. The non-JSON output format is not stable and you will regret relying on it.
-- Supports stdin pipe input for scripting.
-
-Media type auto-detected by extension:
-- Image: `.png` `.jpg` `.jpeg` `.gif` `.webp` `.bmp`
-- Video: `.mp4` `.mov` `.webm` `.mkv` `.avi`
-- File: everything else
+- `--to` defaults to the bound user from the openclaw account
+- `--json` always use this
+- Media type auto-detected: image (.png .jpg .jpeg .gif .webp .bmp), video (.mp4 .mov .webm .mkv .avi), file (everything else)
 
-## Output
+Output:
 
 ```json
 {"ok":true,"to":"user@im.wechat","clientId":"..."}
-```
-
-```json
 {"ok":false,"error":"ret=-2 (rate limited, try again later)"}
 ```
 
-**Exit code 0 means the CLI ran successfully. It does NOT mean the message was delivered.** Always check the `ok` field. Yes, this violates the spirit of Unix exit codes. Reality doesn't care about your principles -- network calls and local execution are different things.
-
-## Errors
+| ret | meaning | action |
+|-----|---------|--------|
+| -2 | rate limited | wait 5-10s, retry |
+| -14 | session expired | re-login via openclaw |
 
-| ret | meaning | what to do |
-|-----|---------|------------|
-| -2 | rate limited | Wait 5-10 seconds and retry. Do not `while true` loop this. That makes it worse. |
-| -14 | session expired | Re-login via openclaw. Tokens don't last forever. |
-
-Rate limit: **~7 messages / 5 minutes** per bot account, enforced server-side, shared across all clients. No, you can't bypass it with multiple processes. Don't try.
+Rate limit: **~7 msgs / 5 min** per bot account, server-side.
 
 ## Accounts
 
 ```bash
 wxclawbot accounts --json
 ```
 
-Auto-discovers from `~/.openclaw/openclaw-weixin/accounts/`.
-
-Env var override (for CI/CD or containers):
+Auto-discovers from `~/.openclaw/openclaw-weixin/accounts/`. Env override:
 
 ```bash
 export WXCLAW_TOKEN="bot@im.bot:your-token"
@@ -149,8 +115,6 @@ export WXCLAW_BASE_URL="https://ilinkai.weixin.qq.com"
 
 ## Programmatic API
 
-Don't want the CLI? Use it as a library:
-
 ```typescript
 import { WxClawClient } from "@herai/wxclawbot-cli";
 import { resolveAccount } from "@herai/wxclawbot-cli/accounts";
@@ -166,15 +130,13 @@ await client.sendText("user@im.wechat", "Hello");
 await client.sendFile("user@im.wechat", "./photo.jpg", { text: "Check this" });
 ```
 
-`resolveAccount()` follows the same discovery logic as the CLI: env vars > local account files.
-
 ## Links
 
 - [npm](https://www.npmjs.com/package/@herai/wxclawbot-cli)
 - [GitHub](https://github.com/lroolle/wxclawbot-cli)
 - [ClawHub](https://clawhub.ai/lroolle/wxclawbot-send) -- `clawhub install wxclawbot-send`
-- [WeChat Integration Tutorial](https://mp.weixin.qq.com/s/nYDQ1obQEHe1WavGpNzasQ) -- start here for basic setup
+- [WeChat Integration Tutorial](https://mp.weixin.qq.com/s/nYDQ1obQEHe1WavGpNzasQ) -- start here
 
 ## License
 
-MIT -- take it and run.
+MIT