The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents standard CLI workflows for database management, including prisma db execute for raw SQL, prisma migrate for schema changes, and prisma dev for local database instances. These commands are essential for ORM functionality and involve expected file system and network interactions.
[DYNAMIC_EXECUTION]: Documents the runtime execution of local project files, such as database seed scripts (tsx prisma/seed.ts) and shell completion generation (source <(prisma complete zsh)). These are standard features for development tooling.
[INDIRECT_PROMPT_INJECTION]: The skill identifies tools with ingestion surfaces, such as prisma db pull (database introspection) and prisma migrate diff (reading schemas from URLs or databases).
Ingestion points: Untrusted data may enter via remote database schemas or SQL files executed through stdin.
Boundary markers: The skill explicitly defines an AI safety checkpoint in references/agent-safety.md, requiring agents to explain impacts and obtain explicit user consent before destructive actions.
Capability inventory: The skill documents capabilities for database read/write, local file creation, and network access for cloud database provisioning.
Sanitization: The skill relies on Prisma's internal validation (prisma validate) and documented user-consent requirements for safety.
[SAFE]: All documented external resources, including documentation links and package references, target official vendor domains (prisma.io, prisma.com) or the vendor's GitHub organization (github.com/prisma).

prisma-cli