magicpath

Pass

Audited by Gen Agent Trust Hub on May 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and execute code using npx magicpath-ai and the magicpath-ai add command, which fetches UI components and their dependencies from the npm registry and MagicPath servers.
  • [REMOTE_CODE_EXECUTION]: The magicpath-ai add command installs React/TypeScript components and potentially new npm dependencies into the local project environment at src/components/magicpath/.
  • [DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command syntax in SKILL.md to execute magicpath-ai info --json at skill load time to provide the agent with immediate project and authentication context. This is used for discovery and is limited to the vendor's own CLI tool.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes theme definitions from the MagicPath API which include a prompt field containing natural-language styling instructions. The agent is explicitly instructed to follow these instructions when adapting components, creating a vulnerability surface for indirect prompt injection.
  • Ingestion points: Theme definitions fetched via magicpath-ai get-theme <id-or-name> -o json and component source code from magicpath-ai inspect <generatedName> -o json.
  • Boundary markers: None. There are no delimiters or instructions provided to the agent to disregard malicious commands embedded in the theme instructions or component source code.
  • Capability inventory: The agent has access to Bash(magicpath-ai *) for network and CLI operations, as well as the ability to read and modify local project files.
  • Sanitization: None. The instructions do not mention validating or sanitizing the input from the prompt field or component source before processing.
  • [COMMAND_EXECUTION]: The skill relies on the execution of the magicpath-ai CLI tool for all core functionalities, including searching, inspecting, and installing components.
Audit Metadata
Risk Level
SAFE
Analyzed
May 12, 2026, 04:33 AM