Skip to main content
By default, evals, enrichments, and actions run at the session level. Use the scope option to target subagent logs individually.

​
Scope options

ScopeRuns at session levelRuns at subagent level
'session' (default)YesNo
'subagent'NoYes
'both'YesYes
// Session only (default)
app.eval('turn-count', ({ stats }) => ({
  pass: stats.turnCount <= 50,
  score: Math.max(0, 1 - stats.turnCount / 100),
}));

// Subagent only
app.eval('agent-efficiency', ({ stats }) => ({
  pass: stats.turnCount <= 10,
  score: Math.max(0, 1 - stats.turnCount / 20),
  message: `${stats.turnCount} turns`,
}), { scope: 'subagent' });

// Both
app.eval('quality-check', ({ stats, source }) => ({
  pass: stats.toolCallCount <= 20,
  score: Math.max(0, 1 - stats.toolCallCount / 40),
  message: `${source}: ${stats.toolCallCount} tool calls`,
}), { scope: 'both' });

​
Subagent context fields

When running at subagent level, the context includes additional fields:
FieldTypeDescription
sourcestring"agent-{id}" - matches entry._source directly
subagentTypestringe.g. 'Explore', 'Bash'
subagentDescriptionstringShort description of what the subagent was doing
parentSessionIdstringThe parent session’s ID
app.eval('adaptive-eval', (ctx) => {
  if (ctx.source !== 'session') {
    console.log(ctx.source);              // e.g. 'agent-a1b2c3'
    console.log(ctx.subagentType);        // e.g. 'Explore'
    console.log(ctx.subagentDescription); // e.g. 'Search for auth code'
    console.log(ctx.parentSessionId);
  }
  return { pass: true };
}, { scope: 'both' });

​
Filtering entries by source

Subagent-scoped evals receive the full combined data (session + all subagents), not just the subagent’s own entries. Use _source to filter to only the current subagent’s entries: 
app.eval('explore-thoroughness', ({ entries, source }) => {
  const myEntries = entries.filter(e => e._source === source);
  return {
    pass: myEntries.length > 5,
    score: Math.min(myEntries.length / 20, 1),
    message: `${myEntries.length} entries`,
  };
}, { scope: 'subagent', subagentType: 'Explore' });
source in the context directly matches the _source value on entries, so the filter is always exact.

​
subagentType filtering

When you specify subagentType, the eval only runs for subagents of that type. Subagents of other types won’t see the eval panel at all. 
// Only runs for Explore subagents
app.eval('explore-depth', ({ entries, source }) => {
  const myEntries = entries.filter(e => e._source === source);
  return {
    pass: myEntries.length > 5,
    score: Math.min(myEntries.length / 20, 1),
  };
}, { scope: 'subagent', subagentType: 'Explore' });

// Runs for all subagent types
app.eval('agent-turn-count', ({ stats }) => ({
  pass: stats.turnCount <= 10,
  score: Math.max(0, 1 - stats.turnCount / 20),
}), { scope: 'subagent' });

​
Subagent enrichments and actions

The same scope and subagentType options apply to enrichments and actions: 
app.enrich('agent-summary',
  ({ stats, entries, source }) => ({
    'Agent Turns': stats.turnCount,
    'Agent Tool Calls': stats.toolCallCount,
    'Agent Entries': entries.filter(e => e._source === source).length,
  }),
  { scope: 'subagent' }
);

app.action('agent-report', ({ entries, source, stats }) => {
  const myEntries = entries.filter(e => e._source === source);
  return {
    output: [
      `Source: ${source}`,
      `Entries: ${myEntries.length}`,
      `Turns: ${stats.turnCount}`,
    ].join('\n'),
    status: 'success',
  };
}, { scope: 'subagent' });