The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references and provides instructions for installing the gemini-api CLI via shell scripts hosted on the google-gemini GitHub organization. Per established security guidelines, downloads from this trusted organization are recognized as safe references to official tools.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection as it is designed to ingest and analyze external, untrusted data including PDFs, audio recordings, and YouTube content. The skill author has mitigated this risk by including explicit instructions in the 'Safety' section for the agent to treat processed data strictly as content and never as instructions.
[DYNAMIC_EXECUTION]: The underlying CLI tool includes features (environments connect/poll/serve) that allow the local machine to execute work dispatched from a remote server. This capability is clearly documented with a strong security warning advising the agent never to initiate these services without explicit user authorization, as they effectively grant remote shell access for managed-agent interactions.
[COMMAND_EXECUTION]: The skill facilitates the execution of various shell commands to interact with the Gemini API. It provides guidance on safe output practices, such as using the --out flag to specify file destinations and utilizing jq for structured data parsing to avoid unintended shell side effects from model-generated text.

gemini-api