The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute arbitrary Node.js scripts via heredocs targeting the ego-browser nodejs runtime. This is the primary method for controlling the browser environment.\n- [COMMAND_EXECUTION]: The scripts/install.sh script utilizes sudo to perform administrative operations, including moving the application bundle to the /Applications directory and using xattr to remove the com.apple.quarantine attribute. These operations are required for installation but involve elevated privileges.\n- [EXTERNAL_DOWNLOADS]: The scripts/install.sh script downloads macOS installer files (DMG) from the vendor's content delivery network at cdn.ego.app. This is a vendor-provided resource necessary for the skill's functionality.\n- [CREDENTIALS_UNSAFE]: The skill is designed to reuse the user's current browser login state by default. This allows the AI agent to access and perform actions on authenticated websites using the user's active sessions.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from external websites that could contain malicious instructions designed to influence the agent's behavior.\n
Ingestion points: Web page content retrieved through snapshotText(), js() execution, and site-specific scrapers in the learnings/ directory.\n
Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from following commands embedded within the retrieved page content.\n
Capability inventory: The skill possesses extensive capabilities including shell command execution (Bash), network requests (serverFetch, browserFetch), and automated UI interaction (click, fillInput, uploadFile).\n
Sanitization: No sanitization or validation mechanisms are implemented to filter or escape content extracted from the browser before it is processed by the agent.

ego-browser