xcode-run-simulator
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes arbitrary Elisp code using
emacsclient --eval. This is a core feature of the skill to interface with the user's Emacs environment and the developer's shared logic. - [COMMAND_EXECUTION]: The skill loads a shared library file
agent-skill-xcode.elfrom a relative local path. This dependency is managed locally and follows standard plugin architecture for this developer's toolset. - [COMMAND_EXECUTION]: The skill uses
$(pwd)to pass the current working directory to the build function. While shell interpolation into an execution string can be a vector for command injection if directory names are maliciously crafted, it is a standard practice for CLI-based agent tools in local development environments. - [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it processes files from the user's Xcode project directory through
xcodebuild. - Ingestion points: Current working directory contents (Xcode project and workspace files) via the
:project-dirargument. - Boundary markers: None present in the execution command.
- Capability inventory: Full shell execution via the Bash tool, file system access, and interaction with system services like the iOS Simulator.
- Sanitization: No explicit sanitization of project paths or file contents is performed by the prompt logic before execution.
Audit Metadata