Most Slack integrations put a bot token in an MCP config file, where every process that can read the config can read the token. Here the credential lives in an encrypted local store and is fetched at send time, and the project states there is no subprocess, exec or shell usage in the path. Mock mode means you can wire up the workflow before you hand it a real token.
A Slack sender built around a local credential vault. The webhook URL, bot token and default channel are stored in MGC Blackbox and read at send time, so the agent that posts the message never holds the secret and nothing sensitive lands in a config file or a log line.
- A Slack message posted with the message body, and optionally a target channel, a custom bot display name and an emoji icon — `send_slack_message`, taking `text`, `channel`, `username` and `icon_emoji`
- Either a webhook URL or a bot token as the delivery path, whichever you stored
- A mock mode for development that exercises the same path without making a real Slack request
Slack credentials — a webhook URL or a bot token — plus a channel, stored in the local vault rather than passed to the server. Install the vault with `pip install mgc-blackbox`, then save a JSON file holding `webhook_url`, `bot_token` and `default_channel` with `mgc_save info_type=config info_owner=slack_bot`; the two identifiers are yours to choose and are how the credentials are retrieved later. Clone the skill and start it with `python mcp_server.py`.