The Agent Skills Directory

[COMMAND_EXECUTION]: The documentation provides an example configuration for chromiumFlags that includes 'disable-web-security': true. This flag disables the Same-Origin Policy (SOP), which is a fundamental security mechanism in web environments. Disabling it can lead to Cross-Site Scripting (XSS) and data exfiltration if the application loads untrusted content.
[COMMAND_EXECUTION]: The skill provides instructions for users to run xattr -cr on macOS. This command removes extended attributes from a file, effectively bypassing macOS Gatekeeper's quarantine flag for unsigned or unnotarized applications, which is a common technique for executing untrusted binaries.
[PROMPT_INJECTION]: The skill describes handling deep links via the open-url event, which represents an indirect prompt injection surface.
Ingestion points: The open-url event handler in SKILL.md receives external URL strings from the operating system.
Boundary markers: The documentation recommends using setNavigationRules to allowlist targets, but the code examples do not explicitly demonstrate prompt-level boundaries or "ignore instructions" warnings.
Capability inventory: The Electrobun framework allows for window management, RPC calls, and file system access (through views:// and file:// protocols).
Sanitization: The skill includes a 'Security Checklist' that specifically recommends routing data through RPC with validation and using navigation rules to mitigate risks from untrusted URLs.

Electrobun Platform