PluginBench
Skill
Pass
Audit score 90

convex-monitor

get-convex/agent-skills

Watch for the next error or request in a Convex app and react to it in real-time.

What is convex-monitor?

Blocks on the next typed event (dev error, prod error, feature request, or heartbeat) in a Convex app instead of polling. Returns the first event to fire, enabling agents to react immediately to changes in local logs, deployments, or production errors.

  • Block on the next dev/prod error, feature request, or quiet heartbeat event
  • Race local error logs, deployment subscriptions, and Sentinel prod-error rows simultaneously
  • Return typed, versioned events with consistent schema across blocking and poll-loop modes
  • Fall back to polling when blocking MCP is unavailable
  • Triage and react to production errors via Sentinel integration

How to install convex-monitor

npx skills add https://github.com/get-convex/agent-skills --skill convex-monitor
Prerequisites
  • Convex project directory with dev or prod app
  • For prod_error events: deployed cloud app and Sentinel access
Claude Code
Cursor
Windsurf
Cline

How to use convex-monitor

  1. 1.Call wait_for_event with project_dir, event_kinds, and timeout_ms
  2. 2.Receive a typed event (convex_error, prod_error, feature_request, or quiet)
  3. 3.Decode the event and take action: fix errors, triage prod issues, or build features
  4. 4.On quiet heartbeat, loop back to step 1

Use cases

Good for
  • Automatically detect and fix the next error in a Convex dev environment
  • Monitor production errors in real-time and trigger fixes without manual polling
  • React to feature requests as they arrive in a deployed Convex app
  • Build an agent loop that responds to app events with minimal latency
  • Combine error detection with automated debugging workflows
Who it's for
  • Convex app developers
  • AI agents automating Convex app maintenance
  • DevOps engineers monitoring production Convex deployments
  • Teams building self-healing or self-improving Convex applications

convex-monitor FAQ

What happens if no event fires before timeout?

The function returns a quiet heartbeat event, allowing the agent to loop or take other actions.

Do I need Sentinel for this skill to work?

Sentinel is only required for prod_error events. Dev errors and feature requests work without it.

What's the difference between blocking and polling modes?

Blocking mode waits for events via MCP; polling mode loops with the same event contract. The event schema and behavior are identical.

Can I filter which event kinds to watch for?

Yes, pass the desired event_kinds in the wait_for_event call to focus on specific event types.

Is the event schema stable across versions?

Yes, the event schema is fixed and versioned, so the same trigger always yields the same typed event.

Full instructions (SKILL.md)

Source of truth, from get-convex/agent-skills.


name: convex-monitor description: "Watch for the next dev/prod error or request in a Convex app and react to it."

<!-- GENERATED from convex-agents content/capabilities/monitor.json — do not edit by hand. -->

Watch for the next thing to react to

Block on the next typed event instead of polling. Races local error logs, deployment subscriptions, and Sentinel prod-error rows; returns the first to fire (or a quiet heartbeat).

Workflow

  1. Call wait_for_event with {project_dir, event_kinds, timeout_ms}.
  2. On kind=convex_error/next_error: decode and fix it. On kind=prod_error: triage (see sentinel) and fix. On kind=feature_request: build it. On kind=quiet: loop.
  3. Where a harness has no blocking MCP (e.g. Copilot cloud), the pack runs a poll loop with the SAME event contract — same behavior, different mechanism.

Rules

  • Prefer the blocking tool; fall back to a poll loop only where blocking MCP is weak.
  • The event schema is fixed and versioned — the same trigger yields the same typed event.
  • Prod events (kind=prod_error) require a deployed cloud app plus Sentinel.