electron

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute Bash commands to launch various native desktop applications (e.g., Slack, Visual Studio Code, Discord, Figma, Notion, Spotify) with the --remote-debugging-port flag. This action exposes the internal state of these applications to external control tools via the Chrome DevTools Protocol (CDP).
  • [DATA_EXFILTRATION]: The skill enables the extraction of potentially sensitive information from automated applications. It provides patterns for saving application state to JSON files (agent-browser snapshot --json > app-state.json), reading text from UI elements (agent-browser get text), and capturing screenshots. This capability can be used to access private communications or sensitive data within apps like 1Password or Slack.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data directly from the user interface of external applications.
  • Ingestion points: Data enters the agent's context through UI snapshots and text extraction commands specified in SKILL.md.
  • Boundary markers: The skill lacks instructions for using delimiters or warnings to ignore embedded commands within the ingested application content.
  • Capability inventory: The agent possesses the capability to simulate user input (clicks, keyboard events, form filling) and execute shell commands through the allowed agent-browser tool.
  • Sanitization: There is no implementation of sanitization or validation for the content retrieved from external applications before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 06:38 AM