The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a robust interface for interacting with the local system through the gologin-agent-browser CLI. It facilitates tasks such as profile creation (profile-create), session control (open, close), and browser interaction (click, type). These commands are essential for the skill's primary purpose of local browser automation.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection due to its interaction with external web content.
Ingestion points: Data enters the agent's context through tools like snapshot and eval which process content from external websites (e.g., workflows/reddit-session-routine.md, workflows/linkedin-leadgen.md).
Boundary markers: The instructions do not define specific delimiters or security warnings to prevent the agent from obeying instructions found within processed web pages.
Capability inventory: The agent can perform significant actions including click, type, and eval across multiple scripts (tools.md), which could be leveraged if an injection occurs.
Sanitization: There is no evidence of sanitization or content filtering applied to the data retrieved from browser snapshots.
[DYNAMIC_EXECUTION]: The skill includes an eval tool that allows for the execution of arbitrary JavaScript code within the browser session (tools.md). This is a standard but powerful capability intended for inspecting page state and performing complex interactions.
[EXTERNAL_DOWNLOADS]: The skill references and encourages the installation of the gologin-agent-browser-cli package and associated GitHub repositories (README.md, references/command-map.md). These are official resources provided by the vendor (gologinlabs) and are consistent with the skill's stated purpose.

gologin-local-agent-browser-skill