Guide · Agent skill
How to use the Convex Event Watcher agent skill (convex-monitor)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The convex-monitor skill teaches an AI agent to stop polling for problems in a Convex app, a hosted backend platform for web apps. Instead, the agent waits on a single blocking tool that returns the first relevant event: a local Convex error, a production error recorded by Sentinel, a feature request, or a quiet heartbeat. This guide explains what the skill does, how to install it, how to trigger it, and what to check before you rely on it.
Convex Event Watcheropenclaw/clawhub · MIT Download the skill Quick start
- 1Download the ZIP. Get the convex-monitor skill ZIP from the directory page.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use, or into your project's .claude/skills/ folder.
- 3Or upload in the Claude apps. Open Settings, then Capabilities, then Skills, and upload the ZIP.
- 4Set up your Convex project. Make sure your Convex project has a working local dev environment.
- 5Confirm your agent harness. Use an MCP-capable harness for the blocking wait_for_event tool, or a poll-loop fallback such as Copilot cloud.
- 6Ask the agent to watch. Request monitoring in plain words, for example asking it to watch for the next error and fix it.
- 7Review each change. Check what the agent does for each event before accepting its edits.
What is an agent skill?
An agent skill is a folder containing a SKILL.md file that gives an AI agent specific instructions for a task. When the agent loads a skill, it follows that workflow for matching requests. This skill ships as a single SKILL.md file inside the .agents/skills/convex-monitor folder of the openclaw/clawhub project, and it is licensed under MIT.
What does the convex-monitor skill make the agent do?
Instead of repeatedly checking for problems, the agent calls a blocking tool named wait_for_event with project_dir, event_kinds and timeout_ms. The call races several sources: local Convex error logs, deployment subscriptions, and Sentinel production error rows. It returns whichever fires first, or a quiet heartbeat if the timeout passes.
- kind=convex_error (next_error): the agent decodes and fixes the local error.
- kind=prod_error: the agent triages the production issue using the Sentinel guidance, then fixes it.
- kind=feature_request: the agent builds the requested feature.
- kind=quiet: the agent simply waits again.
Install the convex-monitor skill
Download the ZIP of the skill from the directory page and unzip it. For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder. Alternatively, in the Claude apps, open Settings, then Capabilities, then Skills, and upload the ZIP there.
Check that the SKILL.md file sits inside the convex-monitor folder after unzipping, so the agent can find it.
- Personal: ~/.claude/skills/
- Project: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
What do you need before it works?
You need a Convex project with a local dev environment. You also need an agent harness that supports blocking MCP tools so wait_for_event can block. If your harness does not, the skill falls back to a poll loop that uses the same event contract, so behavior is the same even though the mechanism differs.
Production events are a separate requirement. The prod_error kind requires a deployed Convex cloud app plus Sentinel. Without both, you will only get local and feature events.
- A Convex project with a local dev environment
- An MCP-capable agent harness, or a poll-loop fallback such as Copilot cloud
- For production triage: a deployed Convex cloud app and Sentinel
How do you trigger the skill with natural requests?
The skill is meant to run as a watcher, so ask the agent to monitor the project and react to what happens next. The source does not list fixed trigger phrases, so describe the intent in plain words. Requests along these lines fit the workflow:
- "Watch my Convex app for the next error and fix it."
- "Monitor production for Sentinel errors and triage any that come in."
- "Keep watching for feature requests and build them as they arrive."
- "Wait for the next event in this project, and loop if nothing happens."
Safety notes: scripts and network
This skill does not run scripts. It does use the network, because it watches deployment subscriptions and Sentinel rows. Review what the agent changes, since the workflow is built to fix errors and build features as they arrive, and the agent may act on noisy or incomplete signals.
Do not use this skill if you want unreviewed automatic changes to your code. Keep a version control checkpoint before a long watch session so you can roll back.
- Runs scripts: no
- Uses network: yes
- Acts on incoming events, so review its changes
Tips for getting good results
Set event_kinds to the sources you actually need, so the agent does not react to events you do not care about. Choose a timeout_ms that suits your session, since a quiet heartbeat simply loops the watch. Use the skill on a Convex project whose local dev environment is running, so local errors can surface.
- Narrow event_kinds to the sources you want.
- Use a sensible timeout_ms for your session.
- Keep a git checkpoint before long watch sessions.
What people use it for
- Catching and fixing a local Convex error as soon as it appears during development.
- Triaging a production error recorded by Sentinel in a deployed Convex cloud app.
- Picking up feature requests as they arrive and building them.
- Keeping an agent loop running with quiet heartbeats when nothing needs attention.
Frequently asked questions
How do I install the convex-monitor skill?
Download the ZIP from the directory page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.
Is the convex-monitor skill free?
The skill is licensed under MIT. The source does not state pricing for Convex, Sentinel or any agent harness, so check those services directly.
What do I need to use convex-monitor?
You need a Convex project with a local dev environment and an agent harness that supports blocking MCP tools, or a poll-loop fallback. Production error handling also requires a deployed Convex cloud app and Sentinel.
Why is my agent not reacting to production errors?
The prod_error events require both a deployed Convex cloud app and Sentinel. If either is missing, production events will not fire, so check both before troubleshooting the skill itself.
Does convex-monitor run scripts on my machine?
No. The skill is listed as running no scripts, but it does use the network to watch deployment subscriptions and Sentinel rows.
Can I use this skill with a project that does not use Convex?
No. The event sources and tools are specific to Convex, so the skill is not suited to projects that do not use it.
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 convex-monitor if you build or maintain a Convex app and want your agent to react to local errors, Sentinel production errors and feature requests as they happen, especially if you can review each change. Pick something else if your project does not use Convex, if you lack a deployed app and Sentinel for production triage, or if you want the agent to make unreviewed changes automatically.
Download the skill