team-workspace
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The installation and startup scripts, such as
scripts/install.mjsandscripts/launcher.mjs, utilize Node.jschild_processfunctions to execute system commands includingPlistBuddy,plutil,xcrun,git, andcodesign. These utilities are used to configure the local application environment, verify system dependencies, and ensure the security of the compiled binaries. - [EXTERNAL_DOWNLOADS]: The
scripts/install.mjsscript downloads and updates plugin components from the author's official GitHub repository (TestAny-io/testany-agent-skills.git) using the Codex CLI's marketplace functionality. - [INDIRECT_PROMPT_INJECTION]: The 'Writing Assistant' feature utilizes the
teamdesk_local_sourcestool to allow the AI to search and read local file content from the user's home directory (e.g.,~/Documentsor~/Projects) to assist in drafting team documentation. While this enables processing of untrusted local data, the sub-agent is restricted by system instructions to only generate text suggestions and is prohibited from executing commands or performing unauthorized network operations. - [DYNAMIC_EXECUTION]: During the installation process, the skill compiles Swift source code located in
assets/launcher/into native macOS binaries. This compilation is required to create the application bundle used to launch the workspace and facilitate communication between the local server and the AI desktop application.
Audit Metadata