astryx
Pass
Audited by Gen Agent Trust Hub on Aug 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches multiple Node.js packages from the NPM registry, including @astryxdesign/core, @astryxdesign/theme-neutral, @astryxdesign/cli, and various StyleX compiler plugins like @stylexswc/nextjs-plugin.
- [COMMAND_EXECUTION]: Utilizes a CLI tool (npx astryx) to perform file system operations, such as initializing projects, injecting page templates into local paths, and running codemods via the 'upgrade' command to modify existing source code.
- [REMOTE_CODE_EXECUTION]: Recommends connecting to a hosted Model Context Protocol (MCP) server at https://astryx.atmeta.com/mcp. This grants the remote server the ability to provide tools and data that can influence agent decision-making and execution.
- [PERSISTENCE_MECHANISMS]: Includes functionality to write persistent behavioral rules to global agent directories, such as ~/.cursor/rules/xds.mdc, which can modify the behavior of AI agents across different user projects and sessions.
- [INDIRECT_PROMPT_INJECTION]: The skill scaffolds instruction files (CLAUDE.md, .cursorrules) for AI agents. These files act as a channel for behavioral instructions. Ingestion Points: Discovery of external packages/components via 'npx astryx discover'. Boundary Markers: None mentioned in generated context. Capability Inventory: CLI file-system operations and MCP server integration. Sanitization: No explicit content filtering mentioned for external data.
- [DYNAMIC_EXECUTION]: The CLI generates React code and layouts from compressed expressions (XLE/XLO) and compiles theme source files into production CSS and JS.
Audit Metadata