Mission Squad Docs

Appearance

Prompt Studio ​

Purpose: Iterate quickly on prompts and verify tool/RAG behavior before publishing or scheduling.

Controls ​

  • Prompt name (optional).
  • Target Model or Agent — pick a named model or an agent.
  • Load prompt from agent — import an existing agent’s prompt for editing.
  • Save prompt to agent — overwrite/save back (requires selecting the agent).
  • Timeout (s) — server execution timeout.
  • Streaming — toggle ON to see SSE deltas and Tool Events live (helpful for debugging).
  • Hide Tool Events — toggle if you only want the final answer.

Usage pattern ​

  1. Select the agent you’re working on.
  2. Toggle Streaming ON and Show Tool Events to watch tool_start/tool_end traces.
  3. Iterate: adjust the system prompt and re‑run until the trace and answer quality meet your criteria.
  4. Save prompt back to the agent.

API parity ​

  • Chat run: POST /v1/chat/completions (OpenAI‑compatible; model may be a model name, a published agent name, or shared/<ownerUsername>/<slug>)
  • Workflows test: POST /v1/core/agent-workflow
  • Streaming uses OpenAI‑style SSE chunks
    See Chat Completions and Core Utilities