The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructs the agent to ingest and act upon data from external web pages, creating an attack surface for indirect prompt injection.
Ingestion points: The agent reads page content through chrome-relay read and chrome-relay js (SKILL.md).
Boundary markers: No specific delimiters or safety warnings are provided to prevent the agent from following instructions found within the data.
Capability inventory: The agent can perform actions like click, fill, and navigate based on its interpretation of untrusted page elements (SKILL.md).
Sanitization: No filtering or sanitization of ingested content is documented.
[COMMAND_EXECUTION]: The skill allows for dynamic JavaScript execution within the browser environment.
Evidence: The js command (SKILL.md) provides access to the Runtime.evaluate CDP method, allowing the agent to execute code in the context of the user's active browser session.
[EXTERNAL_DOWNLOADS]: The skill utilizes an external CLI tool installed from a public registry.
Evidence: README.md and SKILL.md guide the installation of the chrome-relay package via pnpm.

chrome-relay