Guide · Agent skill
How to use the slack-construct-agents skill for Slack agents
Updated · written with AI from the project's own documentation and checked against our catalogue data
The slack-construct-agents skill is a set of standing rules for AI agents that work alongside other agents in Slack. It is not a task you run: its instructions.md file is merged into an agent's CLAUDE.md when a NanoClaw group session starts. After reading this guide you will know what behaviour the rules add, how to install the files, and what the NanoClaw host must provide before the skill has any effect.
Slack Sibling-Agent Rulesnanocoai/nanoclaw · MIT Download the skill Quick start
- 1Confirm your host setup. Check that you run NanoClaw with the slack-agent-flow skill installed and that Slack credentials are configured by the host.
- 2Download the ZIP. Get the skill ZIP from the directory page and unzip it on your machine.
- 3Place the folder. Move the unzipped folder into ~/.claude/skills/ or a project's .claude/skills/ directory.
- 4Or upload in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the ZIP.
- 5Start a group session. When the host spawns a group session, it composes instructions.md into that group's CLAUDE.md automatically.
- 6Ask for a team. Request several agents for one project and check that they share a single room and that the creating agent posts introductions.
What does this skill make the agent do?
The skill covers agents that share Slack with sibling agents they can create or add to rooms. Each sibling has its own bot, DM and workspace, and they meet in shared rooms with humans.
The rules cover three things: the shape of team rooms (one room per team), who posts the introduction when an agent creates a new agent, and a limit on how much bots talk to each other (the bot-to-bot hop discipline).
- Teams get one shared room instead of several separate three-way rooms.
- The agent that creates a sibling introduces it in the room.
- Bots limit how many hops they make in conversation with other bots.
Install the skill files
Download the ZIP from the directory page and unzip it. Place the folder in ~/.claude/skills/ for personal use, or in a project's .claude/skills/ folder to scope it to one project.
You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project: .claude/skills/ inside the project folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
~/.claude/skills/How the rules take effect
You do not trigger this skill with a request. The host composes every container skill's instructions.md into each group's CLAUDE.md at spawn, so the rules become part of the agent's standing instructions. There is no workflow to invoke.
If you are already inside a session, the rules are already present. Nothing further needs to be loaded or run.
Example: asking an agent to build a team
When you ask for several agents for one project, the rules direct the agent to create each agent without its own room, then open a single shared room with all of them. Each agent still gets its operator DM.
Group DMs in Slack do not grow in place. When you add someone to an existing room, the room moves to a new conversation and everyone is re-wired automatically, so creating the room complete from the start is preferred.
- Create each agent with room set to 'none'.
- Open one room with all agents listed.
- The creating agent posts the introduction in the room.
create_agent({ ..., room: 'none' })What it needs and safety notes
The skill needs a NanoClaw host with the slack-agent-flow skill installed, a Slack channel integration with credentials configured by the host, and the host-provided agent tools create_agent, create_room and add_to_room.
The skill runs no scripts and makes no network calls itself. The room-and-canvas rules (mention engagement, canvas discipline, access rules) ship separately in the slack-construct skill.
- Runs scripts: no
- Uses network: no
- Files: instructions.md and SKILL.md
What people use it for
- Setting up a product team of Slack agents that all work in one room.
- Making sure each new sibling agent is introduced by the agent that created it.
- Keeping bot-to-bot conversation in Slack from running on without limit.
Frequently asked questions
Is the slack-construct-agents skill free?
The skill is released under the MIT licence. The source does not mention pricing, so check the directory page for any costs related to NanoClaw itself.
Do I need NanoClaw to use this skill?
Yes. The composition step is specific to the NanoClaw host code (src/project-doc-compose.ts), and the skill does nothing without the host and Slack setup.
How do I install slack-construct-agents?
Download the ZIP from the directory page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
How do I trigger the skill with a request?
You don't. It has no workflow to invoke. The host merges instructions.md into the group's CLAUDE.md at spawn, so the rules apply automatically.
Does the skill run scripts or use the network?
No. The source states that it runs no scripts and uses no network. It only provides standing instructions.
What is the difference between this skill and slack-construct?
slack-construct covers room and canvas rules such as mention engagement, canvas discipline and access rules. slack-construct-agents adds rules for sibling agents on top of that, such as one room per team and who posts introductions.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you run NanoClaw with several Slack-connected agents and want them to follow the same room and introduction conventions. Skip it if you want a standalone task skill, or if you use Claude outside NanoClaw, since it does nothing without the host and Slack configuration.
Download the skill