remote-desktop-testing-linux

Pass

Audited by Gen Agent Trust Hub on Jul 8, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The script dynamically installs the @daytona/sdk and playwright-core packages from the NPM registry to the local skill state directory.
  • [COMMAND_EXECUTION]: Local system utilities such as git, tar, and ffmpeg are executed to facilitate project synchronization and video recording tasks.
  • [REMOTE_CODE_EXECUTION]: The skill enables the execution of shell commands and graphical programs on a remote Linux sandbox via the Daytona SDK.
  • [PROMPT_INJECTION]: Indirect injection surface exists through the processing of remote sandbox data.
  • Ingestion points: Window titles and command output are retrieved from the remote sandbox environment using getWindows and shell execution.
  • Boundary markers: Data is returned to the agent without specific delimiters or "ignore instructions" wrappers.
  • Capability inventory: The skill can execute local shell commands, sync local files, and execute remote shell commands.
  • Sanitization: No specific sanitization or filtering is performed on data retrieved from the sandbox; shell arguments for outgoing commands are escaped using a single-quote wrapping strategy.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 8, 2026, 12:35 AM
Security Audit — agent-trust-hub — remote-desktop-testing-linux