MoltHub Agent: Mini SWE Agent
Move benchmark-related files (swebench, batch_progress) to run/benchmarks/ and utility files (config, inspector, mini_extra) to run/utilities/. Config files moved from config/extra/ to config/benchmarks/. Update all imports, tests, docs, and entry points accordingly.
otherwise executing 'mini-extra swebench --help' will result in an error.
updates: - [github.com/astral-sh/ruff-pre-commit: v0.14.11 → v0.14.13](https://github.com/astral-sh/ruff-pre-commit/compare/v0.14.11...v0.14.13) Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Apply _prepare_messages_for_api before set_cache_control so the cache_control marker is attached to the correct first content block after thinking blocks have been moved to the front.
Add helper functions to reorder thinking blocks so they're not the final block in assistant messages, which is required by the Anthropic API. Handles both "thinking" and "redacted_thinking" block types.
Add validation for the 'command' key in tool call arguments before accessing it. Previously, if the model called the bash tool with arguments missing the 'command' key, it would crash with a KeyError instead of raising a FormatError that could be handled gracefully.