The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install components using npx @agent-native/skills. This package is a part of the official toolset provided by the skill's author (builderio).
[REMOTE_CODE_EXECUTION]: The skill uses npx @agent-native/core@latest to perform local operations such as plan verification and bridge serving. As these are vendor-owned tools, the execution is considered part of the intended functionality.
[COMMAND_EXECUTION]: The skill executes local shell commands via the Agent-Native CLI, such as plan local check and plan local serve, to validate and preview MDX artifacts on the user's machine.
[INDIRECT_PROMPT_INJECTION]: The skill processes external text plans and local repository files as source material to generate visual artifacts. This creates a typical ingestion surface for indirect instructions.
Ingestion points: Processes planText input and local file content through tools like create-visual-plan and import-visual-plan-source (SKILL.md).
Boundary markers: The instructions recommend preserving intent but do not mandate strict delimiters or 'ignore' instructions for the source material ingestion (SKILL.md).
Capability inventory: Access to shell execution (npx), file system writes for MDX files, and network publishing via the Plan MCP connector (SKILL.md).
Sanitization: Visual output in the browser is described as being re-sanitized by the renderer (references/wireframe.md).

visual-plan