The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses package runners like bunx, npx, and pnpm dlx to download and execute official Prisma utilities, including @prisma/cli and create-prisma. It also includes a command to install AI agent skills from the official prisma/skills repository. All identified download targets are vendor-owned resources.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a wide range of CLI commands for application lifecycle management. These include building, running, and deploying apps, as well as managing Prisma Postgres databases and environment variables. The commands are standard for the Prisma Platform CLI and are used for their intended administrative purposes.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project data, such as configuration files and dependency manifests, to automate deployment workflows.
Ingestion points: The agent is instructed to read package.json, prisma.compute.ts, and various framework-specific configuration files (e.g., next.config.ts, vite.config.ts, nest-cli.json) across multiple files, including SKILL.md and references/troubleshooting.md.
Boundary markers: The skill provides a "Source-of-Truth Order" and explicit decision trees to guide the agent in interpreting project files, reducing the risk of accidental obedience to instructions embedded within those files.
Capability inventory: The skill has the capability to write to the file system (via init and scaffolding), execute shell commands for builds and deployments, and perform network operations through the Prisma Platform CLI.
Sanitization: The skill includes mandatory safety rules such as env-do-not-leak-secrets and explicit instructions to never print service tokens, full connection URLs, or secret environment variable values in logs or feedback messages.

prisma-compute