The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the zapier-sdk CLI tool to interact with the Zapier platform, performing tasks such as discovering app actions, testing workflows, and managing deployments. These operations are essential to the skill's primary purpose.\n- [DYNAMIC_EXECUTION]: To facilitate workflow creation, the skill generates TypeScript code (workflow.ts) based on user-defined logic. This code is executed for testing purposes via the zapier-sdk --experimental run-durable command within the Zapier infrastructure.\n- [INDIRECT_PROMPT_INJECTION]: The skill converts user natural language into executable workflow code, creating a surface for potential injection.\n
Ingestion points: User natural language requests in Phase 1 define the workflow steps and logic.\n
Boundary markers: The skill implements a confirmation step (Phase 3) where the agent must summarize the build plan and obtain user consent before any code generation occurs.\n
Capability inventory: The generated code can perform authenticated app actions through sdk.runAction calls in the workflow.ts file.\n
Sanitization: The skill explicitly instructs the use of the Zod library in the generated code to validate and sanitize inputs at runtime.

workflows-create