generate-screenshots

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes various CLI tools to manage the screenshot workflow. It uses the Crowdin CLI for inventory and uploads (crowdin string list, crowdin screenshot upload), invokes UI automation frameworks like Playwright (npx playwright test), Maestro (maestro test), and XCUITest (xcodebuild test), and utilizes system utilities for image capture including adb, xcrun, and screencapture.
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions to install the Maestro automation framework by fetching and executing a shell script from its official domain (https://get.maestro.mobile.dev) using curl | bash. This is a documented method for installing the well-known testing tool.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection through its data ingestion process.
  • Ingestion points: Ingests external string data from the Crowdin API via crowdin string list in JSON format, which is then used to generate a screen plan.
  • Boundary markers: The instructions do not define explicit boundary markers or directives to ignore instructions that might be embedded within the source strings retrieved from Crowdin.
  • Capability inventory: The skill has significant capabilities, including executing shell commands, writing configuration files, and generating test scripts (Playwright specs, Maestro flows) based on the ingested data.
  • Sanitization: There is no explicit mention of sanitizing or escaping the content of the strings before they are used to plan screens or potentially incorporated into generated test locators.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 11:14 PM