Claude Agent SDK Part 6: Fixing the Bash Bypass and Understanding Permissions
Closing the checkpointing loophole and getting to know the SDK's permission system.
Closing the checkpointing loophole and getting to know the SDK's permission system.
Adding the ability for the agent to create posts that follow my templates, with the ability to recover from mistakes.
Building context profiles and usage tracking that works with the SDK's design.
Discovering the trade-offs between agency and control when building on the Claude Agent SDK.
Continuing to explore Anthropic's Agent SDK - making tool calls more descriptive and learning about Python introspection.
Learning in public: experimenting with Anthropic's Agent SDK after hearing that anything Claude Code can do, you can do with the SDK.