Changes:
- #1181: Implement claude code slash commands
- #1180: Minor claude code cleanups
- #1179: Normalize tools in execute_prompt loop
- #1177: gpt-5-codex prompt improvements
- #1178: Don't resolve tool references inside execute_prompt incorrectly
- #1175: Add onExp tag for gpt5CodexAlternatePrompt
- #1165: Update WSL setup note in CONTRIBUTING.md
- #1174: Delete execute_prompt tool
- #1173: Adopt isSubAgent tool call hint
- #1171: refactor: remove endpointType from computeEmbeddings calls for consistency
See More
- #1167: add dev device id as traffic filter for exp
- #1170: Only include todo instructions when todo tool enabled
- #1169: use a relative path in edit traces for replays
- #1168: Add command to show output channel from debug tree view
- #1164: mcp: rework how virtual tool embeddings get expanded
- #1166: Remove TodoListToolInstructions references from agent prompts
- #1161: edits: allow oai config and byok api config for edit tools
- #1154: edits: add edit tool 'learning' for byok
- #1163: Use apply_patch exclusively in msbench
- #1150: Renamed chat commands
- #1152: Wrap token check in catch
- #1151: Fix auto-approve functionality in msbench
- #1162: models - update description for
autowhen no auth - #1160: Consume new CAPI package
- #1155: ensure workspace index is disabled for no auth, disable semantic search provider
- #1159: Surface logging interface
- #1146: nes: fix: better handle enforcing max budget
- #1158: altAction: feat: show model-proposed next edit
This list of changes was auto generated.