The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates interaction with Gemini models by executing a bundled Python script (gemini.py) that uses the Google GenAI SDK.
[DATA_EXFILTRATION]: User-provided prompts and contents of local files (via the @file syntax or visualization feature) are sent to Google's generative AI services. This behavior is the intended purpose of the skill and targets a well-known service provider.
[EXTERNAL_DOWNLOADS]: Installation instructions involve downloading core skill files from the author's public GitHub repository and installing the official google-genai Python package.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it is designed to process and analyze external data, including codebase files and user-generated prompts.
Ingestion points: File contents read during the visualize workflow and text passed to analysis commands.
Boundary markers: The instructions include guidance for the agent to challenge recommendations and look for blind spots, though technical delimiters are not enforced in the script itself.
Capability inventory: The skill can execute shell commands through the bundled script and perform network requests to Google APIs.
Sanitization: No explicit sanitization or filtering of ingested file content is performed prior to being sent to the model.

gemini