The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to verify and deploy backend logic. Specifically, the 'SELF-VERIFY RULE' mandates running npx tsc --noEmit for type checking and npx convex dev --once to push code to a deployment environment. These are standard operations for the Convex platform but represent significant agent capabilities.
[EXTERNAL_DOWNLOADS]: The skill requires external dependencies including the @convex-dev/agent package and the convex CLI tool. These originate from well-known service providers associated with the Convex platform.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data and possesses high-privilege capabilities.
Ingestion points: The skill processes plain-English product ideas ('an app where people can...') and technical backend asks provided by the user (as seen in the 'Workflow' section of SKILL.md).
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands in user input are present.
Capability inventory: The skill has the capability to execute shell commands and perform remote code deployments to the Convex cloud via npx convex dev.
Sanitization: No sanitization or validation of the input data before it is used to generate deployment-ready code is described.

convex-design