llm-router
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The script
scripts/check_install_cli.shautomates the global installation and update of theswiftopenai-clipackage usingnpm. This package is retrieved from the official npm registry (npmjs.com), which is an established well-known service for software distribution. - [COMMAND_EXECUTION]: The skill relies on local shell scripts to verify, install, and configure the
swiftopenaienvironment. It then executes shell commands to invoke the CLI tool with user-defined parameters for processing requests. - [INDIRECT_PROMPT_INJECTION]: The skill acts as an interface that passes user-provided prompts to an external agentic CLI, which constitutes an indirect prompt injection surface.
- Ingestion points: User-provided tasks and questions are passed as string arguments to the
swiftopenai agentcommand inSKILL.md. - Boundary markers: User input is encapsulated within double quotes during command execution to distinguish it from the command itself.
- Capability inventory: The target
swiftopenaitool has broad capabilities including MCP (Model Context Protocol) integration, filesystem access, and tool calling. - Sanitization: The skill does not perform pre-processing or sanitization of the user input, relying on the safety guardrails of the underlying LLM provider and the CLI tool.
Audit Metadata