The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute the native-devtools-mcp package from the npm registry using npx -y or pnpm dlx. This tool provides the core functionality for interacting with the desktop environment.
Evidence: Instructions in SKILL.md and README.md to run sh -c "cd /tmp && exec npx -y native-devtools-mcp".
Evidence: Setup command pnpm dlx native-devtools-mcp setup in SKILL.md.
[COMMAND_EXECUTION]: The skill performs shell operations to launch applications and query local network endpoints.
Evidence: Execution of electron . --remote-debugging-port=9222 to start the target app.
Evidence: Use of curl to verify the Chrome DevTools Protocol endpoint at http://127.0.0.1:9222/json/list.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes content from the Electron application being tested.
Ingestion points: Application DOM content and metadata retrieved via the Chrome DevTools Protocol.
Boundary markers: Absent; there are no specific instructions provided to the agent to ignore or isolate instructions that might be embedded in the application's data.
Capability inventory: Shell command execution, network access to localhost, and UI automation (clicks, keyboard input, screenshots) via MCP tools.
Sanitization: Absent; the skill does not include steps to sanitize or validate the content retrieved from the target application.
[PRIVILEGE_ESCALATION]: To perform desktop automation and screenshots, the skill requires high-level system permissions.
Evidence: The skill notes the requirement for macOS Accessibility and Screen Recording permissions in SKILL.md and README.md.
[EXTERNAL_DOWNLOADS]: The skill fetches the native-devtools-mcp package during setup and execution.
Evidence: Use of npx and pnpm dlx to pull the package from the public npm registry.

electron-live-test