Guide · Agent skill
How to use the Messaging the Human agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Messaging the Human is an agent skill for OpenRig projects. It teaches an agent how to reach the right person or channel when a project needs a human decision or update, check that delivery is ready, and write a short brief a person can read on a phone. After following this guide, you will know how to install the skill, how to ask for it in plain language, and what the agent needs in place before it can send anything.
Messaging the Humanmvschwarz/openrig · Apache-2.0 Download the skill Quick start
- 1Download and unzip the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. In the Claude apps, upload it under Settings, Capabilities, Skills.
- 2Confirm OpenRig and the rig CLI. Make sure the OpenRig daemon is running and the rig CLI is installed. The skill cannot message anyone without them.
- 3Check your gateway. Run rig gateway participants --json to confirm at least one human or external channel is registered.
- 4Set your policy. Make sure your project policy says when a human must be contacted. The skill does not decide that for you.
- 5Ask in plain language. Describe the decision or update you need, such as asking for approval before a change or reporting a finished task.
- 6Review the brief and result. Check the brief the agent wrote and the verify result that shows whether delivery was confirmed.
What an agent skill is and what this one does
An agent skill is a folder with a SKILL.md file that an AI agent reads to learn a specific way of working. This skill covers one job: getting a human involved when the project calls for it, and routing the reply back to the right work.
The skill covers finding the right participant, checking that the delivery channel is ready, writing a brief with a recommendation and the choice requested, handling urgent alerts, and recording decisions honestly. It does not decide when a human must be contacted. That comes from the project's own policy, and the skill itself says that installing it does not create an approval gate.
- Finds people and shared channels registered with the OpenRig gateway
- Checks whether a channel is configured, enabled, active and ready
- Writes a short phone-friendly brief with a recommendation and a requested choice
- Sends urgent asks with an alert flag and records decisions made under a person's decision room
- Routes replies back to the thread they came from
Install the messaging-the-human skill
Download the ZIP for this skill from this page. Unzip it into your personal skills folder at ~/.claude/skills/ so it applies to all your projects, or into a project's .claude/skills/ folder so it applies only to that project.
If you use the Claude apps instead, upload the skill under Settings, then Capabilities, then Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings, Capabilities, Skills, then upload
Set up OpenRig and a gateway first
The skill depends on the OpenRig daemon and the rig CLI. Without them, the agent cannot message anyone. You also need a configured gateway with at least one registered human or external channel, such as Slack. The skill uses the network to reach these channels.
Before relying on the skill, confirm that a registered participant shows up when you list them.
- OpenRig daemon running, with the rig CLI installed
- A configured gateway
- At least one registered human or external channel, for example Slack
- Network access, since the skill sends messages through the gateway
rig gateway participants --jsonTrigger the skill with natural requests
You do not need to name the skill to use it. The skill's description says it applies when project policy calls for a human decision or update, when a human delivery is pending or failed, or when a reply must resume the right work. Ask the agent in those terms.
Good requests describe the situation and the outcome you want. The agent will then look up the participant, check readiness, and write the brief.
- Ask the agent to get a decision from the right person before it changes the production config
- Tell the agent to send a status update to the team Slack channel when the migration finishes
- Ask the agent to check whether the failed approval message was delivered and resend it if policy allows
- Tell the agent to route the human's reply back to the task it was about
What the agent does when it sends a message
The agent first lists the registered people and channels, then inspects the chosen human. It uses the returned address rather than a guessed username. Before sending, it checks that the channel is ready. If the state is indeterminate, the agent treats it as not ready and follows the next inspection step rather than changing the connector setup.
The brief is written for a reader on a phone. It states why the matter matters, the recommendation and the main tradeoff, the bounded action if approved, and the choice requested. Updates state the outcome and say 'No action needed.' The message is sent with the queue create command, and --verify is used to check the result.
- Discover participants and inspect the chosen human
- Check readiness: configured, enabled, active, ready, reason and next action
- Write a brief of roughly 100 to 150 words
- Send with rig queue create and --verify
rig queue create --destination <entityId>@external --human-intent decision --summary "<short subject>" --body-file <brief-file> --evidence-ref <durable-evidence> --verify --jsonUrgent alerts and decision rooms
Add the alert flag only when the request is urgent, such as a sign-in, a credential, or an approval that will expire. A question that can wait for the person's next look does not need it. By default, every ask posts to Slack, so leaving out the alert flag does not hide a request from someone who reads Slack.
Some people set up a decision room, which describes what the team may decide for them while they are away. The agent checks for one before asking. It decides only when the person's own words clearly cover the case and no exclusion applies. Otherwise it asks and waits. A recommended option is not a decision, and a permission prompt or approval is never decided this way.
- Use --alert only for time-sensitive asks
- Check the decision room with rig gateway human mode room <entityId>
- Record a room decision with rig gateway human mode decide and say whether it is reversible
rig gateway human mode room <entityId>Safety notes and limits
The skill declares no scripts. It does use the network, because it sends messages through your configured gateway. It does not describe how content is protected in transit or at rest, so do not send sensitive data to channels based on this skill alone.
The skill also limits itself. It warns against turning solvable technical problems into human gates. It treats messages from shared channels as correspondence, so they never approve or close an ask on their own. If policy leaves a material decision ambiguous, the agent is meant to identify who holds the authority rather than guess.
- No scripts are run by this skill
- It uses the network to reach your gateway and channels
- Sensitive data handling is not documented in the skill files
- Messages from shared channels never approve or close an ask
What people use it for
- Asking a named owner to approve an instance update before it is applied, with the recommendation and tradeoff in the brief
- Sending a short status update to a team Slack channel when a long job completes, with 'No action needed' stated
- Escalating a sign-in or expiring credential as an urgent alert so it reaches someone on urgent-only Slack settings
- Checking whether a failed approval message was delivered and handling the failure rather than guessing
- Routing a person's reply back to the thread and task it concerned
Frequently asked questions
Is the messaging-the-human skill free?
The skill is licensed under Apache-2.0. The source files do not describe pricing, so check the project's page for any cost details. It needs an OpenRig daemon, the rig CLI and a configured gateway to work.
What do I need before I can use this skill?
You need the OpenRig daemon with the rig CLI installed, and a configured gateway with at least one registered human or external channel such as Slack. Without that setup, the agent cannot message anyone.
Where do I install the skill?
Unzip the downloaded ZIP into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project. In the Claude apps, upload it under Settings, Capabilities, Skills.
Does this skill decide when the agent must ask a human?
No. The skill says that when a human must be contacted comes from the project's own policy. Installing the skill does not create an approval gate.
Can the agent approve things without asking me?
Only in narrow cases. If you have set up a decision room and the decision is plainly covered by your own words, the agent can record a decision on your behalf and say whether it can be undone. Permission prompts and approvals are never decided this way.
Why isn't my Slack message showing up?
Check that the channel is ready, meaning configured, enabled, active and ready, using the gateway human show command. The skill treats an indeterminate state as not ready and points to the next inspection step.
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
This skill suits developers running multi-agent OpenRig projects who want agents to escalate decisions to people in a consistent, concise format. It is not useful outside OpenRig, and it is not meant for cases where the agent should decide without oversight or for sending sensitive data to channels, since the files do not describe how that content is protected.
Download the skill