The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies extensively on executing shell commands through the vg CLI to perform all operations, including running models, searching the catalog, and managing files.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of external software packages, specifically vibedgames and @openai/codex via NPM, to function. These are documented as necessary dependencies for the CLI interface.
[DATA_EXFILTRATION]: The vg generate upload command enables the transmission of local files to a remote CDN. While this is a core feature for providing reference images to generative models, it represents a potential vector for exfiltrating sensitive local data if the agent is misdirected.
[INDIRECT_PROMPT_INJECTION]: The skill features a significant attack surface for indirect prompt injection as it ingests and processes untrusted data from external sources.
Ingestion points: External data enters the context through vg generate models (catalog search), vg generate docs (documentation search), and the results returned from vg generate run (model outputs).
Boundary markers: The instructions do not specify any delimiters or safety markers to differentiate between system instructions and data retrieved from the CLI tool.
Capability inventory: The agent has the capability to execute shell commands via the vg CLI, write files to the local system using the --download flag, and upload files to external servers using the upload command.
Sanitization: There is no evidence of sanitization or validation of the external content before it is processed by the agent.

generate