opencode-dev

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides bash scripts for environment initialization and configuration, and allows the definition of custom tools that execute shell commands.
  • [EXTERNAL_DOWNLOADS]: The documentation includes examples of using package runners like npx and uvx to fetch and execute external Model Context Protocol (MCP) servers from well-known sources.
  • [REMOTE_CODE_EXECUTION]: The skill enables the execution of remote code through the configuration of MCP servers and custom tools, which are managed via a built-in permissions system.
  • [PROMPT_INJECTION]: The skill defines surfaces for indirect prompt injection by interpolating data into tool commands (e.g., {{query}} in SQL tools), but includes examples of parameter validation to help sanitize inputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 10:17 PM