The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill contains instructions for the user to execute unverified remote code. Specifically, the 'Setup from scratch' section includes the command bash <(curl -sSf https://gogcli.sh/). This pattern fetches a shell script from a remote URL and executes it immediately within the user's terminal session without any integrity checks or prior review of the script's contents.
[COMMAND_EXECUTION]: The skill makes extensive use of the exec tool to run various command-line utilities, including betaclaw, gog, brew, and gcloud. These commands are used to manage Gmail accounts, set up infrastructure, and send emails. When combined with the remote code execution risk, these capabilities could be abused if the downloaded CLI tool is malicious.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and processing email content. Specifically:
Ingestion points: External data enters the agent's context through browser action=get_text on Gmail pages and via Google Pub/Sub inbound messages.
Boundary markers: There are no instructions or delimiters provided to help the agent distinguish between the skill's instructions and potentially malicious instructions embedded within the emails it reads.
Capability inventory: The skill has access to the exec tool (for CLI commands), browser tool, and web_fetch tool.
Sanitization: The skill description does not specify any sanitization or filtering of the email content before it is processed by the agent.

gmail