cloud-dashboard-walkthrough
Pass
Audited by Gen Agent Trust Hub on Jul 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands and CLI tools to manage the development environment.
- Executes
bashscripts (.devcontainer/test-server-on-daytona.sh) to start sandbox servers. - Launches Google Chrome in headless mode with remote debugging enabled.
- Utilizes the
daytonaCLI for managing sandbox preview URLs and executing commands within sandboxes. - [DATA_EXFILTRATION]: The instructions describe how to programmatically access sensitive authentication data for automation purposes.
- Documentation on retrieving organization invitation tokens via API calls (
GET /api/den/v1/org). - Instructions for generating desktop handoff grants (
POST /api/den/v1/auth/desktop-handoff) to facilitate authentication testing. - These operations are performed within the vendor's own application ecosystem (OpenWork) for legitimate testing workflows.
- [PROMPT_INJECTION]: The skill documents an automation surface that processes external data from a web dashboard.
- Ingestion points: Headless Chrome rendering the OpenWork dashboard (
app.openworklabs.com). - Boundary markers: None; uses standard DOM queries via
browser_evalto drive the UI. - Capability inventory: JavaScript execution in the browser context (
browser_eval), shell command execution (daytona exec), and API interactions. - Sanitization: Not applicable for these automation instructions, but represents a surface where the agent could interact with content injected into the dashboard by other users.
Audit Metadata