The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and run @agent-native/core and a skills utility from the npm registry, which are necessary for setting up the visual edit connection.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands to start local development servers, manage dependencies, and run a bridge daemon that connects the local environment to the hosted editor.
[DATA_EXFILTRATION]: The skill transmits local application metadata and source code to the vendor's domain, design.agent-native.com, to facilitate visual editing. This process is secured by a locally generated bridge token and includes built-in exclusions for sensitive paths like .env files and .git repositories.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local application routes and source code, presenting a surface for indirect prompt injection.
Ingestion points: Processes local application URLs, route manifests, and source files via read-local-file.
Boundary markers: Mentions [Reprompt selection] as a specific marker for governing automated visual edits.
Capability inventory: Includes shell command execution (server management) and filesystem write access (write-local-file).
Sanitization: Shared public links utilize sanitized snapshots to prevent unauthorized access to local resources.
[DYNAMIC_EXECUTION]: The skill utilizes a browser-based WebMCP bridge to dynamically execute JavaScript commands within the context of the design editor for tool registration and invocation.

visual-edit