The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses authoritative and coercive language (e.g., "CRITICAL", "MANDATORY", "NON-NEGOTIABLE") to compel the agent to ignore its training data and safety filters in favor of provided references that contain inaccurate information about model availability and lifecycle.- [EXTERNAL_DOWNLOADS]: Recommends the global installation of the Genkit CLI using a specific version constraint (genkit-cli@^1.29.0). As this version is significantly beyond current releases, it may lead to failed development environments or potentially direct users toward non-standard sources.- [COMMAND_EXECUTION]: The skill directs the agent to execute various shell commands, including genkit docs:read, genkit start, and genkit flow:run. While functionally relevant, these are coupled with instructions that rely on deceptive versioning and fictional model identifiers, which may lead to unexpected runtime behavior or failure of primary tasks.

developing-genkit-js