Agent Transcript Resumption

What: Agents can now resume from previous execution transcripts, allowing you to continue where a previous agent left off.

How to use:

# The Agent tool now accepts a 'resume' parameter
# When invoking an agent programmatically:
{
  "subagent_type": "investigator",
  "prompt": "Continue analyzing the codebase",
  "resume": "agent-abc123"  // ID of previous agent execution
}

Details:

• New resume parameter in Agent tool input schema
• Loads previous agent's message chain from transcript file
• Filters messages by agent ID and reconstructs conversation history
• Prepends previous messages to new prompt for continuity
• Previous messages are not re-recorded to session storage
• Evidence: New function b_2() at line 465741; resume parameter schema at line 449027-449032; resume logic at lines 449114-449130