The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute the bdg CLI utility. This utility manages the lifecycle of browser sessions, including starting instances, capturing screenshots, and forcefully terminating Chrome processes during cleanup operations.
[CREDENTIALS_UNSAFE]: The tool includes a specific command (bdg network getCookies) to retrieve session cookies from the browser. This provides a direct path for the agent to access sensitive session identifiers or credentials stored in the browser.
[DATA_EXFILTRATION]: Through network inspection commands like bdg network headers and bdg network list, the agent can view sensitive information such as authentication headers (e.g., Bearer tokens) and response payloads.
[PROMPT_INJECTION]: The skill processes untrusted data originating from the browser, which represents an indirect prompt injection surface.
Ingestion points: Content is ingested through console logs (bdg console), network request/response data (bdg network list), and DOM tree inspection (bdg dom get).
Boundary markers: No explicit boundary markers or delimiters are defined to separate untrusted browser content from the agent's instructions.
Capability inventory: The skill is configured to use the Bash tool (for CLI execution) and the Read tool (for filesystem access).
Sanitization: There is no evidence of sanitization or filtering of the data retrieved from the Chrome DevTools Protocol before it is processed by the agent.

bdg-browser-debug