Google is retiring Gems, its custom prompt feature for Gemini chat, and replacing them with Skills. The transition marks a strategic shift toward agent-ready prompt architectures that multiple AI labs now converge around.
Skills function as detailed, reusable prompts that users invoke manually by typing "/" or that Gemini executes automatically based on context. Unlike Gems, which were essentially custom chatbot personalities, Skills adopt an open standard developed by Anthropic. This standardization matters because it allows prompts to work across different AI platforms and enables autonomous agents to call and chain Skills together without user intervention.
Google will phase out Gems starting in November, with existing Gems migrating automatically to Skills format. The company has not announced a hard deadline for Gems removal, but the transition appears permanent.
The move reflects a broader industry trend. OpenAI rolled out GPTs earlier, which share similar DNA with Skills. Anthropic published the open standard that Google now adopts. Anthropic has also integrated Skills into Claude, positioning them as native building blocks for agentic workflows. The convergence around Skills format suggests the AI industry views this architecture as essential infrastructure for the next phase of AI products.
Skills differ from Gems in three key ways. First, they use structured, standardized formatting rather than natural language instructions. Second, they're designed to be chainable and composable. An autonomous agent can call multiple Skills in sequence to complete complex tasks. Third, they embed metadata and constraints that let systems understand what a Skill does, what inputs it requires, and what outputs it produces. This machine-readable structure enables reliable automation.
The timing aligns with Google's broader push toward agentic AI. The company has emphasized agents as the next frontier in AI development, following foundation models. Agents need standardized ways to access tools and prompts. Skills provide that interface.
Anthropic's involvement matters. The company published the open standard that Skills use, positioning itself as a standards setter rather than just a product company. This move reflects a play for long-term influence in the AI ecosystem. OpenAI and Google adopting similar approaches validates the standard and locks in Anthropic's fingerprint on how the industry builds agents.
For users, the transition from Gems to Skills means better portability and automation potential. A Skill built for Gemini could eventually run on Claude or other platforms. More importantly, as AI agents become the primary interface for AI products, Skills will serve as the primitives that agents compose into workflows.
The shift also signals that free-form, natural language prompt engineering is becoming legacy infrastructure. The industry is standardizing around structured, machine-readable formats that humans and agents both understand. This reduces ambiguity and enables reliable delegation to autonomous systems.
Google has not announced pricing changes or new Skills-specific features. The migration appears designed primarily for compatibility and interoperability rather than new capabilities. However, the standardization opens doors for future development, including cross-platform Skills marketplaces and more sophisticated agent-to-agent interactions.