The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the use of the @agent-native/core package via npx for installation and bridge management, which involves downloading and executing remote code from the npm registry.
[COMMAND_EXECUTION]: Instructs the agent to execute shell commands to start a local bridge daemon (design connect) and discover application routes, utilizing environment variables for secure token handling.
[DATA_EXFILTRATION]: The skill facilitates the reading of local source files and application manifests to be displayed in a hosted Design MCP environment. It includes built-in safeguards that exclude sensitive paths such as .env and .git from being accessed or transmitted.
[INDIRECT_PROMPT_INJECTION]: The skill processes live content from a running local application, creating an attack surface for indirect prompt injection.
Ingestion points: Discovery of application routes and live iframe content served through the bridge.
Boundary markers: Implements a [Reprompt selection] hard write boundary for proposed changes.
Capability inventory: Includes capabilities for read-local-file, write-local-file, and apply-visual-edit to modify project source code.
Sanitization: Requires mandatory human write consent for filesystem modifications and uses version-checked transactions to prevent conflicting edits.

visual-edit