This skill defines the Discord control boundary for a Hermes/OpenClaw agent.
The goal is not to let Discord run arbitrary shell commands. The goal is to give a private channel a tiny, reviewable command surface:
- ask for status,
- summarize the inbox,
- draft triage decisions,
- apply approved decisions,
- log what happened.
Install
Install the curated public source files into your local skills directory.
Configure
Create a Discord application and bot. Install it only into the private server where the agent should operate.
Set these environment variables on the VPS:
Optional when the bot will inspect Linear:
Commands
The v1 command surface should stay small:
Safety rules
- One private Discord channel.
- Explicit allowed users or roles.
- No broad message-reading behavior.
- No raw prompt passthrough.
- No arbitrary shell execution.
- No mutation without a reviewable decision ID.
- No public Hermes/OpenClaw port.
Run the preflight check
The included check confirms that the six environment variables used by the starter adapter are present before you register slash commands or start the Discord adapter.
It reports which checked values are missing, prints the expected command surface, and exits with an error when any checked value is absent. It does not verify credentials, Discord IDs, allowlist syntax, or the Hermes command itself. After it passes, follow the implementation checklist in SKILL.md to register the slash commands and connect the adapter to Hermes.
Course
Use this with the Always-On Agents lesson:
The Discord channel should exist before the X bookmark sync starts creating items that need review.