The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses npx to execute vendor-provided CLI tools such as @prisma/cli and create-db. These commands are used to discover available regions, provision databases, and inspect command help as part of the intended setup process.
[EXTERNAL_DOWNLOADS]: The skill instructions include fetching an additional skill module, prisma-orm-setup, from the official Prisma GitHub repository (github.com/prisma/skills) if it is not found locally. This is a standard modular design for vendor skills.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes untrusted user-controlled data during project inspection.
Ingestion points: Reads project configuration files (package.json, environment files), environment variables, and MCP tool schemas.
Boundary markers: Explicitly instructs the agent to preserve unrelated environment entries and avoid logging sensitive data, which helps mitigate accidental data disclosure.
Capability inventory: Includes file read/write operations for environment configuration, network access via vendor SDKs, and command execution via npx.
Sanitization: Instructions mandate that the agent must not print or log secrets, tokens, or connection strings in the chat or logs.

prisma-postgres-setup