convex-deploy-guard
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill implements safety protocols for deployment management. It instructs the agent to identify and announce deployment targets (local, dev, preview, prod) before executing any modifications. It enforces strict consent requirements for production environments and provides guidance on using official Convex-specific CLI tools and MCP flags responsibly.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a surface for processing data from local configuration files to determine deployment targets. This behavior is necessary for the skill's primary purpose of deployment safety.
- Ingestion points: Workflow step 1 (reads .env.local and convex.json).
- Boundary markers: Absent; instructions do not specify using delimiters when reading project configuration.
- Capability inventory: Deployment commands (npx convex deploy), code execution (npx convex run), environment variable modification (env set), and Convex MCP tool access.
- Sanitization: Absent; the skill relies on the integrity of local project files.
- [COMMAND_EXECUTION]: The skill guides the agent in running npx convex commands for deployment and environment management. These commands are vendor-provided and execution is protected by mandatory identification and user consent steps.
Audit Metadata