superlevels-chrome-extension

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to download the source code from a repository on GitHub (levelsio/superlevels).
  • [COMMAND_EXECUTION]: Provides instructions to execute shell commands including git clone and cd to set up the development environment.
  • [DYNAMIC_EXECUTION]: The Live CSS Editor feature utilizes chrome.scripting.executeScript to inject and apply user-provided CSS styles directly into the active browser tab's context at runtime.
  • [DATA_EXFILTRATION]: The Music Recognizer feature captures audio streams from the current tab using chrome.tabCapture and transmits the data to the ACRCloud API for identification. This operation requires the user to manually provide their own API credentials in the settings.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where it processes external data, including user-supplied CSS and webpage DOM elements for GDPR banner dismissal. It also contains self-referential instructions in the 'Security Audit' section that guide the analyzing agent on how to evaluate the skill's safety.
  • Ingestion points: User input from the CSS editor textarea; DOM selectors on third-party websites for consent banner identification.
  • Boundary markers: No delimiters or 'ignore embedded instructions' warnings are present to isolate untrusted data.
  • Capability inventory: Uses chrome.scripting.executeScript for style injection and chrome.tabs.remove for tab management.
  • Sanitization: Employs textContent instead of innerHTML when injecting CSS, which provides a layer of protection against basic script injection within the style element.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 10:36 AM