x-composer
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes several local scripts and system tools to perform automation tasks. It runs
scripts/cdp-launch.js,scripts/cdp-type.js, andscripts/x-post.shvia the shell, and utilizes system utilities includingopen(to launch Chrome),pbcopy(to manipulate the clipboard), andosascript(to control browser focus and simulate pasting).\n- [DYNAMIC_EXECUTION]: The skill generates and executes code at runtime to handle browser interactions and data formatting. In Playwright mode, it usesbrowser_run_codeto execute arbitrary JavaScript within the browser context. Thex-post.shscript dynamically generates a Python command string to convert text input into JSON segments.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and post user-supplied text, which introduces a potential vector for instructions embedded in the content. Ingestion points include arguments tox-post.shand input typed into the browser. While no specific delimiters are used, the skill mandates a human-in-the-loop confirmation step before posting. Capabilities include file writing to the home directory and full browser DOM control.\n- [PRIVILEGE_ESCALATION]: The skill uses AppleScript (osascript) to simulate system-level keystrokes (Cmd+V) and activate applications. This type of interaction typically requires specific OS-level permissions (Accessibility) to allow the agent to control the user interface across application boundaries.
Audit Metadata