ai-dvps/comate/send-wecom-msg

<objective> Send WeCom messages to any user using the wecom send CLI command (@webank/wecom). Automatically detects structured content and uses markdown formatting when appropriate. Helps draft messages with formatting guidance before sending. </objective>

<quick_start> Send a message to a user:

wecom send --to-user USERID --message "Hello from the team" --session-id ${COMATE_SESSION_ID}

Send a markdown message:

wecom send --to-user USERID --message "**Bold** and `code`" --msg-type markdown --session-id ${COMATE_SESSION_ID}

If wecom is not in PATH, use npx wecom or the full path from WECOM_CLI_PATH. </quick_start>

<workflow>

1. Verify CLI version: Before any send operation, check the installed CLI version:

``bash wecom --version ` Expected: 1.2.0 or higher. If the version is lower, advise the user to update: `bash npm install -g @webank/wecom@latest ` If wecom is not found, check npx wecom --version or ${WECOM_CLI_PATH} --version. If the CLI is not installed at all, advise the user to install it: `bash npm install -g @webank/wecom ``

1. Determine intent: Is the user asking to send immediately or draft first?
2. Extract recipient: Identify the target user ID from the user's request
3. Build the message:

• If the user provides a message, use it as-is
• If drafting, help format the message and confirm before sending

1. Detect markdown: If the message contains any of these patterns, use --msg-type markdown:

• **bold**, *italic*, ~~strikethrough~~
• ` code or ` code blocks ``
• - list items or 1. numbered lists
• [links](url)
• > quotes
• # headings
• Tables with | separators

1. Get session ID: Use ${COMATE_SESSION_ID} for the current session
2. Execute the command:

``bash wecom send --to-user <USERID> --message "<MESSAGE>" [--msg-type markdown] --session-id ${COMATE_SESSION_ID} ` If wecom` is not available in PATH, try:

• npx wecom send ...
• ${WECOM_CLI_PATH} send ...

1. Report result: Show the command executed and the outcome

</workflow>

<examples> <example number="1"> <input>Send a message to ZhangSan saying the deployment is complete</input> <output>

wecom send --to-user ZhangSan --message "The deployment is complete" --session-id ${COMATE_SESSION_ID}

</output> </example>

<example number="2"> <input>Notify LiSi that the build failed with error logs</input> <output> Draft message (auto-detected markdown):

**Build Failed**
 
Error:

TypeError: Cannot read property 'id' of undefined at processData (src/utils.ts:42)

 
Please check the logs.

Send with:

wecom send --to-user LiSi --message "**Build Failed**\n\nError:\n\`\`\`\nTypeError: Cannot read property 'id' of undefined\n    at processData (src/utils.ts:42)\n\`\`\`\n\nPlease check the logs." --msg-type markdown --session-id ${COMATE_SESSION_ID}

</output> </example>

<example number="3"> <input>Send a greeting to WangWu</input> <output>

wecom send --to-user WangWu --message "Hello WangWu!" --session-id ${COMATE_SESSION_ID}

</output> </example> </examples>

<anti_patterns> <pitfall name="guessing_recipient"> Never guess a user ID. If the recipient is unclear, ask: "Who should receive this message? Provide the WeCom user ID." </pitfall>

<pitfall name="missing_quotes"> Always wrap the message in double quotes. Unquoted messages break when the shell interprets special characters. </pitfall>

<pitfall name="ignoring_exit_code"> The wecom CLI returns specific exit codes:

• 0: Success
• 1: Invalid arguments or context file error
• 2: No WeCom bot context found
• 3: HTTP request failed

Report the actual exit code and meaning to the user. </pitfall> </anti_patterns>

<success_criteria>

• Recipient is specified via --to-user
• Message is properly quoted for shell execution
• Markdown is auto-detected when appropriate
• Session ID is passed via --session-id
• The actual CLI command is shown before or during execution
• Results (success or error) are reported clearly

</success_criteria>