The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to interact with the host system, primarily to launch and manage applications during test execution (e.g., using open -a).
[COMMAND_EXECUTION]: The documentation in known-issues.md suggests using dangerouslyDisableSandbox: true as a workaround for application execution issues. This recommendation encourages bypassing security isolation mechanisms intended to protect the host environment.
[PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection by analyzing untrusted local data to determine its testing logic.
Ingestion points: The agent identifies application features by exploring the local codebase (Phase 2) and reads existing test definitions from test-cases/**/*.md.
Boundary markers: The instructions do not implement boundary markers or instructions to disregard potential commands embedded within the analyzed project files.
Capability inventory: The skill possesses high-privilege capabilities, including the ability to execute shell commands (Bash), modify files (Write, Edit), and control the GUI via the Computer Use MCP.
Sanitization: There is no evidence of filtering or validation for the content ingested from the codebase before it influences the agent's scenario generation or execution steps.

computer-use-test