The Agent Skills Directory

[DATA_EXFILTRATION]: The skill provides the command actionbook browser cookies list, which allows the agent to extract all cookies from the current browser session, including sensitive session tokens and authentication data.
[DATA_EXFILTRATION]: The 'Extension mode' (--extension flag) allows the agent to control the user's actual Chrome browser, granting access to live sessions, existing logins, and private data across all open tabs.
[COMMAND_EXECUTION]: The actionbook browser eval command permits the execution of arbitrary JavaScript within the browser's context, which could be used to manipulate page content, steal local storage data, or perform actions on behalf of the user.
[PROMPT_INJECTION]: The skill is highly vulnerable to Indirect Prompt Injection. It ingests untrusted data from external websites using actionbook browser text, html, and snapshot commands without explicit boundary markers or sanitization, potentially allowing malicious web content to influence agent behavior.
Ingestion points: Web content is retrieved via browser text, browser html, and browser snapshot in SKILL.md and command-reference.md.
Boundary markers: None identified; instructions do not specify the use of delimiters or 'ignore' instructions for scraped content.
Capability inventory: The agent can perform file writes (screenshots/PDFs), execute JavaScript (eval), and perform network-based navigation.
Sanitization: No evidence of input validation or content filtering before passing data to the LLM.
[COMMAND_EXECUTION]: The actionbook extension install command performs system-level modifications by registering a native messaging host on the local machine, which is a form of software installation that alters browser and OS configurations.

actionbook