The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes existing workflow source code (workflow.ts) and metadata, creating a surface for indirect prompt injection.
Ingestion points: Workflow source files and metadata fetched via get-workflow-version and get-workflow-draft commands.
Boundary markers: The instructions provide guidelines to "apply the requested change narrowly," but the skill lacks formal delimiters to isolate external code from the agent's instructional context.
Capability inventory: The skill uses zapier-sdk to list, fetch, create, update, publish, and run durable workflows, along with jq for file processing.
Sanitization: There is no explicit sanitization or filtering of the fetched source code before it is presented to the agent for analysis and modification.
[COMMAND_EXECUTION]: The skill heavily relies on executing shell commands via the zapier-sdk CLI to interact with the Zapier platform. This includes fetching secrets (connection IDs), reading source code, and publishing executable workflows.
[COMMAND_EXECUTION]: The skill uses jq to dynamically assemble JSON objects from local files for use in subsequent CLI commands.
[EXTERNAL_DOWNLOADS]: The skill mentions a compatibility gate that may require downloading tools from the zapier/agent-skills GitHub repository. These are official vendor resources and are used for environment validation.

workflows-modify