slack-cli-read（Slack CLI read 搜索）

slack-cli-read

Use this skill to read Slack content via the global slack command.

Scope

• Read-only commands are supported.
• Writing/sending/replying is not implemented yet. If asked to write, state that it is not available yet and offer to draft message text only.

Environment

• Assume slack is available as a global command.
• Slack credentials are expected at ~/.config/slack/slack.env.
• Required variables:

- SLACK_WORKSPACE - TOKEN - D_COOKIE

• If auth fails or env vars are missing, help the user populate that file.

Command Map

• Login bootstrap (interactive): slack auth login <workspace>
• Auth diagnostics: slack auth status
• Verify auth/workspace: slack me
• List all conversations: slack chat list
• List unread conversations: slack chat list --unread
• List DM inbox: slack dm list
• List unread DMs: slack dm list --unread
• Read chat history: slack chat history "<chat>" --limit 30
• Read DM history: slack dm history <user_or_dm_id> --limit 30
• Read full thread: slack thread show <chat_or_dm> <thread_ts>
• Read one full message: slack chat message <chat_or_dm> <ts>
• Search users: slack users list --query <text>
• Raw API call (expert mode): slack api call <endpoint> -p key=value
• Raw API curl wrapper (expert mode): slack api curl <endpoint> -- [curl args]

Working Style

1. Start broad with chat list or dm list when intent is vague.
2. Narrow into history with a bounded --limit.
3. Expand parent threads via thread show when inline replies are truncated.
4. Fetch exact messages with chat message when full text is needed.

Command Hygiene

• Quote names starting with # in shell commands, for example: slack chat history "#general".
• Prefer IDs when names are ambiguous.
• Keep commands bounded with --limit unless the user asks for more.
• Use pretty output for human reading.
• Use --format json or --format jsonl for structured parsing.
• Use slack api call/slack api curl only when typed commands do not cover the needed endpoint behavior.
• If auth fails, run slack auth status before retrying read commands.

Response Guidance

• Summarize key findings instead of dumping raw command output.
• Keep timestamps and IDs in results when they help the user drill down.
• Treat attachment lines (📎...) as message context.