skills/zapier/sdk/zapier-sdk/Gen Agent Trust Hub

zapier-sdk

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the installation of official Node.js packages from the @zapier scope (@zapier/zapier-sdk, @zapier/zapier-sdk-cli, @zapier/zapier-durable) and standard development dependencies (typescript, zod, @types/node) from the npm registry.
  • [COMMAND_EXECUTION]: Extensive use of the npx zapier-sdk CLI tool is documented for discovery, authentication, and execution of actions across integrated applications.
  • [REMOTE_CODE_EXECUTION]: The skill documents the publish-workflow-version and run-durable commands, which allow users to upload local TypeScript source code to Zapier's cloud infrastructure (code-substrate-runner) for execution as durable workflows. This is a primary feature of the SDK.
  • [INDIRECT_PROMPT_INJECTION]: The skill interacts with over 9,000 third-party applications via Zapier. Data retrieved from these external sources (e.g., email bodies, CRM records, chat messages) is ingested into the agent's context, presenting a surface for indirect prompt injection.
  • Ingestion points: Data returned from zapier.runAction, zapier.listActions, and zapier.fetch documented in SKILL.md and references/workflows.md.
  • Boundary markers: The documentation encourages the use of Zod (InputSchema) for structured data validation in references/examples.md and references/workflows.md.
  • Capability inventory: Subprocess execution (npx), file writes/deployment (publish-workflow-version), and network operations (zapier.fetch) across multiple scripts.
  • Sanitization: Validation of input shapes via Zod is demonstrated in example code to ensure data integrity.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 04:08 PM
Security Audit — agent-trust-hub — zapier-sdk