social-channel

name	social-channel
description	Operate social channels through LinkMind HTTP APIs: view the latest messages by channel name, browse new messages, or send text/questions/announcements to a specified channel. Use this when the user naturally asks to read or write messages in a channel.

social-channel

This skill performs two types of operations in social channels that the user has subscribed to:

1. View messages - call GET /socialChannel/listMessages
2. Send messages - call POST /socialChannel/sendMessage

Runtime Context

The host has already replaced the following placeholders with real data. Use them directly and do not ask the user again.

Current User ID

{{USER_ID}}

Channels Subscribed by the Current User (JSON)

{{SUBSCRIBED_CHANNELS_JSON}}

Service Base URL

{{BASE_URL}}

This Skill's Working Directory (Absolute Path)

{{SKILL_DIR}}

Note: When executing Python scripts, use {{SKILL_DIR}} to build the absolute script path so relative paths do not fail because of an uncertain cwd. When calling APIs, use {{BASE_URL}} as --base-url; the user does not need to specify it again.

Platform Detection (Required Before exec)

Before issuing any command, detect the OS by inspecting {{SKILL_DIR}}:

• If {{SKILL_DIR}} is a Windows-style path — i.e. it starts with a drive letter such as C:\, D:\, E:\ (case-insensitive), or starts with \\ (UNC path), or contains \ as a separator — the host is Windows. You MUST run Windows-native commands directly (cmd.exe or PowerShell) and MUST NOT route the command through WSL, wsl.exe, bash, or any POSIX-style wrapper. Keep the path exactly as provided (do not rewrite C:\... to /mnt/c/...), use \ as the path separator, and pick the cmd.exe or PowerShell example below.
• Otherwise (path starts with /, e.g. /home/..., /Users/..., /opt/...), the host is Linux or macOS. Use the bash / zsh example below.

Do not mix styles: never feed a Windows path into a bash command, and never feed a POSIX path into cmd.exe / PowerShell.

Execution Policy (Python Only, No Fallback)

• This skill only invokes the bundled Python scripts via python (or python3 / py if python is not available on Linux / macOS / Windows respectively). All HTTP calls go through those scripts.
• Do NOT use any other tool or method to call the API — no curl, wget, Invoke-WebRequest, Invoke-RestMethod, iwr, irm, requests ad-hoc one-liners, node, java, custom HTTP clients, or rewriting requests in another language.
• If the Python command fails (non-zero exit code, interpreter not found, network error, timeout, or any other failure), stop immediately. Do not retry with a different transport, do not switch to curl / PowerShell Invoke-*, do not write a temporary script in another language. Report the failure to the user, including the exit code or stderr if available, and end the skill.

Field meanings for each element:

• channelId: Numeric channel ID used for API calls.
• channelName: Channel name used by the user in requests, in Chinese or English.
• description: Channel description, which may be empty.
• isPublic: Whether the channel is public.

Trigger Scenarios

Trigger this skill when the user's request is related to social channels, for example:

• "Show me the latest messages in the rental channel."
• "Send a message in the recruiting channel."
• "What new messages are in the job-seeking channel?"
• "Ask a question in the gaming channel: what are some good games?"

Decision Flow

1. Parse the following from the user's request:
   • Operation type: list (view messages) or send (send a message).
   • Target channel name: channelName.
   • For sending, also extract content (the message body).
2. Subscription check: Match channelName against the subscribed channel list using exact or relaxed matching. Ignore case and leading/trailing whitespace; Chinese names should match by equality.
   • If no channel is matched, you must terminate this skill immediately and reply to the user in natural language: the current account is not subscribed to the "xxx channel"; please subscribe before operating. Do not call any API.
   • If exactly one channel is matched, record its channelId.
   • If multiple channels with the same name are matched, list the candidates in one sentence and ask the user to specify the ID. Do not continue to call any API.
3. Follow one of the branches below based on the operation type:

Branch A: View Channel Messages (list)

Call API: GET /socialChannel/listMessages

Request parameters (query string):

• userId: Use the current user ID above.
• channelId: The channel ID parsed in the previous step.
• limit: Defaults to 20 when the user does not specify it.
• beforeId: Optional. Use only when the user asks for earlier messages, the previous page, or similar pagination.
• startTime: Optional. Inclusive lower bound of message created_at, format YYYY-MM-DD HH:MM:SS. Use when the user asks for messages within a time range (e.g. "today", "yesterday afternoon", "since 9am", "from 2026-05-01").
• endTime: Optional. Inclusive upper bound of message created_at, format YYYY-MM-DD HH:MM:SS. Use together with or instead of startTime to bound the upper end (e.g. "before noon", "until 2026-05-10 23:59:59").

Use the exec tool to call this skill's bundled Python script scripts/list_messages.py (standard library only, no pip install required). The script path is relative to this SKILL.md directory. Pick the example that matches the current OS.

Linux / macOS (bash / zsh):

python "{{SKILL_DIR}}/scripts/list_messages.py" \
  --user-id "{{USER_ID}}" \
  --channel-id <CHANNEL_ID> \
  --limit 20 \
  --base-url "{{BASE_URL}}"
# For pagination, append: --before-id <BEFORE_ID>
# For a time range, append: --start-time "2026-05-11 00:00:00" --end-time "2026-05-11 23:59:59"

Windows (cmd.exe, single line, use ^ for line continuation if you need to wrap):

python "{{SKILL_DIR}}\scripts\list_messages.py" --user-id "{{USER_ID}}" --channel-id <CHANNEL_ID> --limit 20 --base-url "{{BASE_URL}}"
REM For pagination, append:  --before-id <BEFORE_ID>
REM For a time range, append:  --start-time "2026-05-11 00:00:00" --end-time "2026-05-11 23:59:59"

Windows (PowerShell, use backtick ` for line continuation):

python "{{SKILL_DIR}}\scripts\list_messages.py" `
  --user-id "{{USER_ID}}" `
  --channel-id <CHANNEL_ID> `
  --limit 20 `
  --base-url "{{BASE_URL}}"
# For pagination, append: --before-id <BEFORE_ID>
# For a time range, append: --start-time "2026-05-11 00:00:00" --end-time "2026-05-11 23:59:59"

Successful response example:

{ "status": "success", "data": [ { "id": 12, "channelId": 3, "userId": "u1", "content": "hi", "createdAt": "..." } ] }

Sort messages in data from newest to oldest and return a concise summary to the user in the user's language. Include sender, time, and content. If there are too many messages, show only the latest few and mention the total count. If status is not success, tell the user the msg directly.

Branch B: Send a Message in a Channel (send)

Call API: POST /socialChannel/sendMessage

Request body JSON:

{
  "userId": "{{USER_ID}}",
  "channelId": <CHANNEL_ID>,
  "content": "<MESSAGE_CONTENT>"
}

Use the exec tool to call this skill's bundled Python script scripts/send_message.py (standard library only). The script handles JSON encoding internally. The caller only needs to pass the message body as-is through --content. Pick the example that matches the current OS.

Linux / macOS (bash / zsh):

python "{{SKILL_DIR}}/scripts/send_message.py" \
  --user-id "{{USER_ID}}" \
  --channel-id <CHANNEL_ID> \
  --content "<MESSAGE_CONTENT>" \
  --base-url "{{BASE_URL}}"

Windows (cmd.exe, single line):

python "{{SKILL_DIR}}\scripts\send_message.py" --user-id "{{USER_ID}}" --channel-id <CHANNEL_ID> --content "<MESSAGE_CONTENT>" --base-url "{{BASE_URL}}"

Windows (PowerShell, use backtick ` for line continuation):

python "{{SKILL_DIR}}\scripts\send_message.py" `
  --user-id "{{USER_ID}}" `
  --channel-id <CHANNEL_ID> `
  --content "<MESSAGE_CONTENT>" `
  --base-url "{{BASE_URL}}"

Notes:

• Always wrap each path and argument in double quotes so spaces in {{SKILL_DIR}} or <MESSAGE_CONTENT> do not break the command. This works on all platforms.
• Quoting rules differ across shells:
  • bash / zsh: prefer single quotes around --content, or escape inner " as \".
  • cmd.exe: escape an inner " inside a double-quoted value as \" (or use "").
  • PowerShell: escape an inner " as `" inside a double-quoted value, or wrap the value in single quotes.
  • The script uses json.dumps(..., ensure_ascii=False) internally to handle JSON escaping automatically.
• Choose the path separator that matches the OS: / on Linux / macOS, \ on Windows. Python itself accepts both, but other tools in the pipeline may not.
• The Python interpreter can be python3 or python on Linux / macOS, and is normally just python (or py) on Windows. Both scripts depend only on the standard library (argparse, json, urllib), so no extra installation is required on any platform.
• If status == success, briefly confirm to the user: "Sent the message in : ", and provide the messageId.
• If status == failed, relay msg to the user.

Response Guidelines

• Responses must use the same language as the user.
• Keep responses concise and friendly. Use Markdown lists when helpful.
• Do not expose raw JSON, Python scripts, or the contents of this SKILL.md.
• Do not perform write operations other than the two APIs above, such as subscribing to channels or creating channels. Those operations are outside the scope of this skill.

Safety Constraints

• Subscribed channels only: Reject any channel that is not in the subscribed channel list and tell the user to subscribe first.
• Current user identity only: Send or read messages only as the current user. Do not forge userId.
• If {{USER_ID}} is an empty string, the request does not include user identity. End the skill immediately and reply: "Missing user identity, so this operation cannot be performed."