Claude Code AGENTS.md Standardization Conflict

Shopify CEO Tobi is considering banning Claude Code internally because it only reads Anthropic's proprietary CLAUDE.md file and ignores the industry-emerging AGENTS.md and .agents/skills conventions. Claude Code representative Thariq responded that the team is working on making Claude Code more customizable, with future support for AGENTS.md. Anthropic's position is that different model families are not interchangeable—system prompts significantly affect performance, and Claude models have specific preferences for skills, system prompts, and CLAUDE.md formatting. A temporary workaround allows referencing AGENTS.md within CLAUDE.md using an @ mention.
Key insights
Proprietary Configuration Formats Create Vendor Lock-in Concerns
Shopify's threat to ban Claude Code highlights growing enterprise frustration with proprietary agent configuration formats. As organizations adopt multiple AI coding tools, the lack of a universal standard like AGENTS.md forces teams to maintain separate configurations for each tool. Anthropic's argument that model-specific formats optimize performance creates tension between tool optimization and workflow standardization.
Model-Specific Optimization Conflicts with Industry Standardization
Anthropic maintains that Claude models have specific preferences for how skills, system prompts, and configuration files are formatted, and that generic formats may sacrifice performance. Claude Code even maintains different system prompts for different models. This technical argument for specialization directly conflicts with the industry's push toward the AGENTS.md standard, creating a trade-off between maximum performance and cross-tool compatibility.