refactor-agent

Prompt file imported from prashanthgopishetty/doctalk (.github/prompts/refactor-agent.prompt.md). Copyright stays with the author.

Analyze and improve the following agent file: $AGENT_FILE_PATH

Please:

1. Read $AGENT_FILE_PATH completely

2. Run through this checklist:

   • Are all async functions using async def?
   • Are all return types annotated?
   • Is the system prompt clear and concise (< 500 chars ideally)?
   • Is the tool call loop handling errors gracefully?
   • Are there any bare except: blocks?
   • Is there any duplicated logic shared with other agents?
   • Does the agent correctly return a state dict (not mutate in place)?
   • Are tools imported from backend/app/tools/ (not hardcoded)?
   • Is the LLM binding done once at module level (not inside the node function)?

3. List all issues found before making changes:

   Issue 1: [description] — Line X
   Issue 2: [description] — Line X
   

4. Apply fixes for all issues found

5. Summarize changes made after editing

Follow .github/instructions/langgraph.instructions.md and .github/agents/self-improvement.agent.md. Do not change the agent's behavior or routing — only improve the code quality.