The Agent Skills Directory

[COMMAND_EXECUTION]: Uses the zapier-sdk CLI to perform administrative tasks on workflows, including listing versions, creating drafts, and publishing updates. These are standard operations for the Zapier platform.
[DYNAMIC_EXECUTION]: Features a synthetic testing capability using zapier-sdk --experimental run-durable to execute workflow code in a controlled environment before deployment. This allows for validation of logic without immediate production impact.
[INDIRECT_PROMPT_INJECTION]: Processes external source code from Zapier workflows.
Ingestion points: Workflow content fetched via get-workflow-version and get-workflow-draft as described in SKILL.md.
Boundary markers: Instructions direct the agent to focus on narrow edits and maintain existing code boundaries and schemas.
Capability inventory: File system access (reading/writing workflow.ts) and shell execution via zapier-sdk.
Sanitization: Utilizes jq to ensure code and configuration data are correctly escaped and formatted as JSON when passed to CLI arguments.
[SAFE]: The skill interacts exclusively with official vendor (Zapier) infrastructure and tools. It implements robust security practices, including optimistic concurrency control (via draft_revision), detailed pre-publish summaries, and mandatory user confirmation checkpoints.

workflows-modify