glm-web-search
Pass
Audited by Gen Agent Trust Hub on Feb 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
npxto download and execute themcporterutility from the npm registry to manage MCP server connections. - [COMMAND_EXECUTION]: Several shell commands are executed to set up the environment, including creating directories with
mkdir, writing configuration files withcat, and using a Python one-liner to parse JSON data. - [PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface because it retrieves untrusted content from the internet through the search tool.
- Ingestion points: Data returned by the
webSearchPrimesearch tool inSKILL.md. - Boundary markers: No specific delimiters or warnings are defined to prevent the agent from executing instructions found in search results.
- Capability inventory: The agent has the ability to make network requests (via the search service) and execute local commands (via the
mcporterCLI). - Sanitization: No sanitization or safety checks are implemented for the external data fetched during searches.
Audit Metadata