The Agent Skills Directory

[COMMAND_EXECUTION]: Multiple Python scripts (daemon_ctl.py, run_with_env.py, check_env.py) utilize the subprocess module to run local development tools like pnpm, node, corepack, and lsof. These operations are limited to a predefined allowlist of actions (e.g., pnpm tools-dev status, corepack enable) necessary for managing the Open Design lifecycle.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to download the Open Design source code from its official GitHub repository (nexu-io/open-design) and install plugins from the soia-team vendor. These are legitimate resources required for the skill's stated purpose.
[DATA_EXPOSURE]: The skill performs local environment discovery by checking for the existence of configuration files (e.g., ~/.claude.json, ~/.codex/config.toml) and probing local network ports on 127.0.0.1. This data is used solely to determine which Open Design integration route (CLI, Desktop, or MCP) is active on the host machine.
[REMOTE_CODE_EXECUTION]: The script install_od_mcp.py automates the registration of Open Design as a Model Context Protocol (MCP) server within other AI agent environments. It achieves this by modifying the JSON or TOML configuration files of those agents to include execution paths for the Open Design Helper application. The script implements security best practices by providing diff previews, creating .bak-od backups, and requiring explicit user confirmation via an --apply flag.
[INDIRECT_PROMPT_INJECTION]: The skill processes external design files (DESIGN.md, tokens.css) and project HTML assets. To mitigate potential instruction injection from these files during synchronization or auditing tasks, scripts/od_sync.py implements a strip_comments function that removes HTML, CSS, and JavaScript comments before scanning content for design rule violations (redlines).

soia-dev-open-design-ops