tabz-terminals
Warn
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The
scripts/setup-worktree.shscript automatically executesnpm install,yarn install, orpnpm installon newly created worktrees. These package managers can execute arbitrary code during installation via lifecycle hooks (such aspostinstall) defined in a project'spackage.jsonfile. - [COMMAND_EXECUTION]: The
scripts/setup-worktree.shscript automatically runsnpm run build, which executes arbitrary commands defined in the project's build configuration. - [COMMAND_EXECUTION]: The skill provides functionality to spawn terminals and execute arbitrary shell commands via a local REST API endpoint (
http://localhost:8129/api/spawn). - [CREDENTIALS_UNSAFE]: The skill reads an authentication token from the local file
/tmp/tabz-auth-tokento authenticate with the TabzChrome API. Exposure or misuse of this token could allow unauthorized control over terminal sessions. - [PROMPT_INJECTION]: The skill orchestrates parallel workers using natural language prompt templates. This architecture creates a surface for indirect prompt injection, where malicious instructions embedded in untrusted repository content could influence the behavior of the workers.
Audit Metadata