The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install or upgrade the official genkit-cli package from the public npm registry. This is a standard prerequisite for using the framework and targets a well-known package from a trusted organization.
[COMMAND_EXECUTION]: The genkit CLI is used to perform routine development tasks, such as searching and reading documentation, executing flows for testing, and launching the developer UI. These commands are used as intended within the Genkit ecosystem.
[COMMAND_EXECUTION]: The automated setup process in references/setup.md involves scanning the workspace, creating directory structures (e.g., ai/tools, ai/flows), and generating configuration files. It also adds a development script to the package.json file. These actions are transparently defined to facilitate project initialization.
[SAFE]: A surface for indirect prompt injection is present as the skill processes user-supplied data in prompts and supports external data grounding via Google Search. This surface is inherent to the functionality of an AI application framework and does not represent a malicious design.
Ingestion points: User-provided inputs in flow definitions (SKILL.md) and examples (references/examples.md), as well as external content from googleSearchRetrieval and documentation searches.
Boundary markers: Delimiters or 'ignore' instructions are not explicitly used in the provided prompt template examples.
Capability inventory: The skill utilizes CLI execution and file system writes during setup and development tasks.
Sanitization: Input validation or sanitization logic is not demonstrated in the minimal reproducible examples.

developing-genkit-js