A built-in skill ships with Claude Code for working directly with the Claude Platform. It handles model migrations, API features like prompt caching, and onboarding to newer APIs like Claude Managed Agents.
Running /claude-api migrate scans your codebase and guides you through updating model names, prompts, and effort settings when moving to a newer model like Opus 4.7.
Running /claude-api managed-agents-onboard walks you through configuring a Claude Managed Agent for your use case and gives the environment the context it needs to monitor and debug it afterward.
The Claude Platform skill is fully open source, covering SDKs in seven languages plus the CLI. It's also integrated into CodeRabbit, JetBrains, Resolve AI, and Warp.
The team is actively tracking down high-friction bugs and is soliciting community input on their worst offenders. If you have a persistent pain point, now is a reasonable time to surface it.
A long-standing bug where Claude Code appears to freeze during large file writes has reportedly been tracked down. The team has been chasing this one for a while.
The apparent hang issue during Claude Code operations doesn't appear universally; it seems to hit certain users disproportionately, while others rarely encounter it.
A member of the Claude team confirmed a fix is underway for an issue reported by a user, with no additional detail shared publicly.
The no-flicker renderer is being polished to become the default, with more updates on the way. A broader sweep of smaller papercut fixes is also in progress alongside it.
A new fullscreen renderer is available via /tui fullscreen, offered as a significantly better experience for users running into display issues.
A rendering bug in the terminal UI may be addressed by the new no-flicker renderer, accessible via the /tui command.
Larger image support is in progress; in the meantime, /clear-images lets users remove images from the conversation context.
A team member flags that pasting images into conversations may be quietly consuming context, raising the question of whether users are aware of that tradeoff.
A team member mused that doing a full clear might be worth it over /compact when it removes a significantly larger share of context.
You can invoke the built-in Claude Platform skill conversationally, for example by asking "Add prompt caching with the Claude API," and it applies best practices automatically without a slash command.
