kelly-ecommerce-intel

Pass

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources such as marketplace pages, competitor pricing, and customer reviews to generate signals and drafts.
  • Ingestion points: External e-commerce websites and news sources browsed by the agent as part of the daily intelligence collection (SKILL.md).
  • Boundary markers: All generated signals, actions, and drafts are held in a Busabase-based review cockpit (AirApp) for human approval before any external handoff or execution.
  • Capability inventory: The skill has the ability to write records to Busabase and perform status updates through administrative scripts like scripts/execute_decisions.mjs.
  • Sanitization: The frontend application includes utility functions to escape HTML characters before rendering data in the cockpit (app.js).
  • [COMMAND_EXECUTION]: The skill includes several Node.js scripts for managing the application lifecycle and executing human decisions.
  • scripts/setup.mjs: Provisions the Busabase workspace (Folders and Bases) required for operation.
  • scripts/publish_airapp.mjs: Handles the deployment of the local cockpit code to the Busabase platform.
  • scripts/execute_decisions.mjs: A CLI tool that processes approved items and marks them as completed in the database.
  • [DYNAMIC_EXECUTION]: The application uses standard ECMAScript dynamic imports for modular loading of data providers and configuration.
  • Evidence: The provider factory (content/kelly-ecommerce-intel-app/app/js/providers/index.js) dynamically imports demo-provider.js or busabase-provider.js based on the URL context.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 9, 2026, 01:36 PM