convex-deploy-guard
Classify + announce the target Convex deployment before any deployment-affecting
by get-convexFree
Description
convex-deploy-guard is an agent skill that stops commands from reaching the wrong Convex deployment. A Convex project can have several deployments, such as a personal development one, a preview and production, often across several projects on the same machine, and the skill's premise is that most incidents start with a command aimed at the wrong one. Before any deployment-affecting command, the skill identifies the target by reading CONVEX_DEPLOYMENT in the local environment file, the convex.json file and whether a deploy key is set, or by calling the official Convex MCP status tool. It then announces the target so the user can see it before anything runs. Production is treated as consent-gated: each production action needs a fresh, explicit yes, and that consent does not carry over to later actions or sessions. The skill also supports a read-only mode for a whole session, which blocks changes to deployments. It is part of the get-convex agent-skills repository and is added to a compatible agent with the skills command. The audits passed with Gen Agent Trust Hub, Socket and Snyk. It suits teams who run several Convex environments and want a safety check before deploys.…
Highlights
- Identifies the deployment before any change
- Announces the target before acting
- Fresh consent for every production action
- Session-wide read-only mode
Ratings & Reviews
No ratings yet
Be the first to rate convex-deploy-guard.
People also visit
deployment-patterns
Deployment workflows, CI/CD pipeline patterns, Docker containerization
convex-deploy-guard FAQ
What does convex-deploy-guard do?
It identifies and announces the Convex deployment a command will affect before the command runs.
How does it identify the target?
From CONVEX_DEPLOYMENT in the environment file, convex.json, whether a deploy key is set, or the Convex MCP status tool.
How is production handled?
Each production action needs a fresh, explicit yes, and consent is per action and per session.
Is there a read-only mode?
Yes, a session can be set to read-only, which blocks changes.
How do I install it?
Run: npx skills add https://github.com/get-convex/agent-skills --skill convex-deploy-guard
What security audits has it passed?
Gen Agent Trust Hub, Socket and Snyk all passed.
What are the best convex-deploy-guard alternatives?
Popular alternatives to convex-deploy-guard include safety-guard, convex-domains and convex-self-heal. See all convex-deploy-guard alternatives