skills/open-gsd/gsd-core/gsd-surface/Gen Agent Trust Hub

gsd-surface

Warn

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill performs dynamic loading of JavaScript modules using computed paths derived from environment variables and the user's home directory. Specifically, it executes code from ${runtimeConfigDir}/gsd-core/bin/lib/surface.cjs and capability-registry.cjs via the require() function.
  • [COMMAND_EXECUTION]: The skill utilizes Bash for path resolution and management of the ~/.claude configuration environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill accepts user-supplied arguments to modify the active 'surface' of the agent, potentially allowing a user or malicious input to enable or disable clusters of skills. 1. Ingestion points: User-provided tokens in the $ARGUMENTS variable within SKILL.md. 2. Boundary markers: No explicit boundary markers or 'ignore' instructions are used to separate user input from system logic. 3. Capability inventory: The skill has access to Read, Write, and Bash tools, along with the ability to dynamically execute local JavaScript code. 4. Sanitization: While cluster names are validated against a list, profile names and general command routing depend on input tokens without comprehensive sanitization.
  • [EXTERNAL_DOWNLOADS]: The documentation references the installation of the @opengsd/gsd-core package from the NPM registry. This package is recognized as a resource provided by the skill's authoring organization.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 7, 2026, 12:43 AM
Security Audit — agent-trust-hub — gsd-surface