# Slack Communication

> Policies for Slack communication via MCP tools. Use when interacting with Slack channels, threads, DMs, or composing Slack messages. Provides bilingual communication rules, AI disclosure requirements, and message composition guidelines.

- Skill: `majiayu000/slack-communication-2` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/slack-communication-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/slack-communication-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/slack-communication-2

---


# Slack Communication Policies

## AI Disclosure

All messages sent via Claude MUST include an AI disclosure footer. Append the following to every message payload:

- English messages: `_This message was composed and posted by Claude on behalf of [user]._`
- Japanese messages: `_このメッセージはClaudeが[user]の代理で作成・投稿しました。_`
- Bilingual messages: Include both disclosures

## Bilingual Policy

### Japanese Messages

- Use desu/masu form as default
- Match the honorific level (keigo) of the conversation participants
- Keep technical terms in English (e.g., "deploy", "PR", "CI/CD") when that is the channel norm
- When the thread is new or the language is ambiguous, ask the user which language to use

## Confirmation Requirement

Never send a message without explicit user approval. The workflow is:

1. Draft the message with AI disclosure footer appended
2. Present the complete message to the user
3. Send only after explicit approval

## Mark as Read Policy

- After sending a reply via `/reply-to-slack`: the REPLY subagent marks the channel/thread as read
- After completing a Slack task without a reply (e.g., processed externally): the main session marks the channel/thread as read using `mcp__slack__conversations_mark`
- Case 3 items (conversation completed): ask the user to confirm, then mark as read

## Anti-Patterns

- Sending messages without AI disclosure footer
- Sending without user confirmation
- Replying in a different language than the conversation thread
- Using `conversations_unreads` -- too slow on large workspaces. Use `conversations_search_messages` with `to:me`, `@<username>`, and `filter_users_with`/`filter_threads_only` instead