The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to read sensitive local files, including the root .env file and .superset/auth-token.enc, to identify ports, API origins, and authentication state.
[COMMAND_EXECUTION]: The skill uses shell commands like lsof to find active debug ports and bun to execute development scripts, start the application, and seed the database.
[DYNAMIC_EXECUTION]: The skill utilizes Runtime.evaluate over the Chrome DevTools Protocol (CDP) to execute arbitrary JavaScript code inside the Electron application's renderer process for testing and automation.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by ingesting and processing data from the running application's UI, DOM state, and console logs, which could potentially contain untrusted content.
Ingestion points: Application UI measurements (e.g., scrollTop), active route information, and console.error logs retrieved from the renderer.
Boundary markers: No specific delimiters or boundary instructions are provided to isolate processed application data from the agent's control flow.
Capability inventory: The agent can execute JavaScript via Runtime.evaluate, run shell scripts via bun, and inspect system networking via lsof.
Sanitization: The skill provides defensive instructions to specifically avoid printing or logging authentication tokens and raw response bodies during sign-in procedures.

cdp-verification