Research a question on the public web

Use this scenario

Run

Run web_research scenario.

Use run_scenario with the saved scenario name and string-valued inputs. Required values are requested before execution.

View invocation JSON

JSON

{
  "scenario_name": "web_research",
  "inputs": {
    "question": "your question",
    "source_count": "5"
  }
}

Watch this version

V9 / Gemma 4 ↗

web_research2 INPUTS / 1 TOOLS / 4 STEPS

Inputs

PARAMETERS

Input	Requirement	Default value
`question`	Required	`No default`
`source_count`	Optional	`5`

Agent instructions

SYSTEM PROMPT

Coordinate saved search, reader and synthesis scenarios. Never invent evidence or save research to user memory.

Tool access

CAPABILITIES

web_check_answer_operation — selected from this version's tool catalog.

Ordered steps

EXECUTION

01
NESTED SCENARIO / FRESH CHILD
web_search ↗
- question = {{question}}
- source_count = {{source_count}}
Save output as search.
02
BOUNDED COLLECTION / FRESH CHILDREN
Read search.results as source. Target {{source_count}}; at most 30 attempts. Collect reading.
Count successful readable results; distinct identity: url.
1. 01
  NESTED SCENARIO / FRESH CHILD
  web_read_source ↗
  - question = {{question}}
  - url = {{source.url}}
  Save output as reading.
Save output as findings.
03
NESTED SCENARIO / FRESH CHILD
web_synthesize ↗
- question = {{question}}
- evidence = {{findings}}
- search_status = {{search.engines}}
Save output as answer.
04
BOUND OPERATION / HOST EXECUTED
web_check_answer_operation
- answer = {{answer}}
- evidence = {{findings}}
Save output as verified_answer.

Return to main

COPY EXIT

Return the verified final answer unchanged, including source links and actual coverage.

The final text returns without model rewriting. Bound operation scenarios return application results; pending changes still require host approval.

Complete XML, including any embedded C#. Copy it without changing the file's contents.

XMLStructureInputsTools / C#StepsExit

<scenario schemaVersion="3" id="web_research" title="Research a question on the public web">
  <inputs><input name="question" required="true"/><input name="source_count" default="5"/></inputs>
  <systemPrompt>Coordinate saved search, reader and synthesis scenarios. Never invent evidence or save research to user memory.</systemPrompt>
  <tools><allow name="web_check_answer_operation"/></tools>
  <steps>
    <run scenario="web_search" result="search"><input name="question" value="{{question}}"/><input name="source_count" value="{{source_count}}"/></run>
    <foreach items="search.results" item="source" collect="reading" result="findings" limit="{{source_count}}" attempts="30" successProperty="readable" identityProperty="url">
      <run scenario="web_read_source" result="reading"><input name="question" value="{{question}}"/><input name="url" value="{{source.url}}"/></run>
    </foreach>
    <run scenario="web_synthesize" result="answer"><input name="question" value="{{question}}"/><input name="evidence" value="{{findings}}"/><input name="search_status" value="{{search.engines}}"/></run>
    <tool name="web_check_answer_operation" result="verified_answer"><input name="answer" value="{{answer}}"/><input name="evidence" value="{{findings}}"/></tool>
  </steps>
  <exit mode="copy">Return the verified final answer unchanged, including source links and actual coverage.</exit>
</scenario>

<scenario schemaVersion="3" id="web_research" title="Research a question on the public web">
  <inputs><input name="question" required="true" /><input name="source_count" default="5" /></inputs>
  <systemPrompt>Coordinate saved search, reader and synthesis scenarios. Never invent evidence or save research to user memory.</systemPrompt>
  <tools><allow name="web_check_answer_operation" /></tools>
  <steps>
    <run scenario="web_search" result="search"><input name="question" value="{{question}}" /><input name="source_count" value="{{source_count}}" /></run>
    <foreach items="search.results" item="source" collect="reading" result="findings" limit="{{source_count}}" attempts="30" successProperty="readable" identityProperty="url">
      <run scenario="web_read_source" result="reading"><input name="question" value="{{question}}" /><input name="url" value="{{source.url}}" /></run>
    </foreach>
    <run scenario="web_synthesize" result="answer"><input name="question" value="{{question}}" /><input name="evidence" value="{{findings}}" /><input name="search_status" value="{{search.engines}}" /></run>
    <tool name="web_check_answer_operation" result="verified_answer"><input name="answer" value="{{answer}}" /><input name="evidence" value="{{findings}}" /></tool>
  </steps>
  <exit mode="copy">Return the verified final answer unchanged, including source links and actual coverage.</exit>
</scenario>

Download XML ↓

EVERY XML ELEMENT / IN FILE ORDER

Every part, briefly.

Short explanations of every element and chat step in this scenario.

01
<scenario>
Names and versions the workflow.
Writing example ↗
02
<inputs>
Declares input values.
Writing example ↗
03
<input>question
Requires a supplied value.
Writing example ↗
04
<input>source_count
Provides an optional value.
Writing example ↗
05
<systemPrompt>
Sets the child's instructions.
Writing example ↗
06
<tools>
Defines available capabilities.
Writing example ↗
07
<allow>web_check_answer_operation
Grants an existing tool.
Writing example ↗
08
<steps>
Orders the chat turns.
Writing example ↗
09
<run>
Invokes a fresh saved child scenario.
Writing example ↗
10
<input>question
Binds a named operation or child input.
Writing example ↗
11
<input>source_count
Binds a named operation or child input.
Writing example ↗
12
<foreach>
Processes and collects a bounded array.
Writing example ↗
13
<run>
Invokes a fresh saved child scenario.
Writing example ↗
14
<input>question
Binds a named operation or child input.
Writing example ↗
15
<input>url
Binds a named operation or child input.
Writing example ↗
16
<run>
Invokes a fresh saved child scenario.
Writing example ↗
17
<input>question
Binds a named operation or child input.
Writing example ↗
18
<input>evidence
Binds a named operation or child input.
Writing example ↗
19
<input>search_status
Binds a named operation or child input.
Writing example ↗
20
<tool>web_check_answer_operation
Executes a declared host operation.
Writing example ↗
21
<input>answer
Binds a named operation or child input.
Writing example ↗
22
<input>evidence
Binds a named operation or child input.
Writing example ↗
23
<exit>
Returns the output unchanged.
Writing example ↗

Write a scenario, step by step ↗How sub-agents work ↗