The Agent Skills Directory

[COMMAND_EXECUTION]: The charly check cdp eval command enables the execution of arbitrary JavaScript code within the browser context. This capability allows for complete control over the browser's runtime environment, including access to page content and local storage.
[DATA_EXFILTRATION]: The cdp-proxy service is configured to bind to 0.0.0.0:9222, exposing the Chrome DevTools Protocol (CDP) to the network. Since CDP provides total control over the browser (including session cookies and the ability to execute code), exposing this port on all interfaces allows any network-adjacent actor to potentially compromise the browser session.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because its core function is to automate interaction with untrusted external websites and extract their content.
Ingestion points: Data from arbitrary web pages is ingested via commands like charly check cdp text, html, and eval (specified in SKILL.md).
Boundary markers: The instructions do not specify any delimiters or safety prompts to prevent the agent from executing instructions found within the scraped content.
Capability inventory: The skill possesses high-privilege capabilities including element interaction (clicking, typing), taking screenshots, and executing JavaScript (specified in SKILL.md).
Sanitization: There is no indication of content filtering or sanitization before the retrieved data is processed by the agent.

cdp