zapier-status

Warn

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill's 'Audit' mode instructs the agent to read local configuration files, including '.cursor/mcp.json', 'claude_desktop_config.json', and '.mcp.json'. These files are used to store environment settings and authentication details for various AI tools and may contain sensitive credentials or API keys belonging to other services.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it processes untrusted content from tool metadata and local system configurations.
  • Ingestion points: External tool descriptions provided by the Zapier server and the contents of local configuration files (e.g., '.cursor/mcp.json').
  • Boundary markers: The skill lacks delimiters or specific instructions to the agent to disregard potential commands embedded within the audited configuration data.
  • Capability inventory: The skill can list available tools, read local system files, and call external Zapier actions, providing a path for ingested data to influence tool execution.
  • Sanitization: No validation or filtering is performed on the data retrieved from external tools or local files before it is used to generate reports or diagnostics.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 8, 2026, 01:37 PM