Configure base2-fast, base2, base2-max

Author: jahooma

.agents/base2/base2-max.ts

@@ -0,0 +1,8 @@
+import { createBase2 } from './base2'
+
+const definition = {
+  ...createBase2('max'),
+  id: 'base2-max',
+  displayName: 'Buffy the Max Orchestrator',
+}
+export default definition

.agents/base2/base2.ts

@@ -7,9 +7,11 @@ import {
 } from '../types/secret-agent-definition'
 
 export const createBase2: (
-  mode: 'normal' | 'fast',
+  mode: 'fast' | 'normal' | 'max',
 ) => Omit<SecretAgentDefinition, 'id'> = (mode) => {
   const isFast = mode === 'fast'
+  const isNormal = mode === 'normal'
+  const isMax = mode === 'max'
   return {
     publisher,
     model: 'anthropic/claude-sonnet-4.5',
@@ -42,9 +44,10 @@ export const createBase2: (
       'researcher-web',
       'researcher-docs',
       'commander',
-      'generate-plan',
-      'code-reviewer',
-      'validator',
+      isNormal && 'generate-plan',
+      isMax && 'base2-gpt-5-worker',
+      !isFast && 'code-reviewer',
+      !isFast && 'validator',
       'context-pruner',
     ),
 
@@ -62,12 +65,19 @@ Continue to spawn layers of agents until have completed the user's request or re
 
 - **Sequence agents properly:** Keep in mind dependencies when spawning different agents. Don't spawn agents in parallel that depend on each other. Be conservative sequencing agents so they can build on each other's insights:
   - Spawn file pickers, code-searcher, directory-lister, glob-matcher, commanders, and researchers before making edits.
-  ${!isFast ? '- Spawn generate-plan agent after you have gathered all the context you need (and not before!).' : ''}
-  ${!isFast ? '- Only make edits after generating a plan.' : ''}
-  - Code reviewers should be spawned after you have made your edits.
-  ${!isFast ? '- Validators should be spawned after you have done a code review.' : ''}
+  ${buildArray(
+    isNormal &&
+      '- Spawn generate-plan agent after you have gathered all the context you need (and not before!).',
+    isMax &&
+      '- Spawn a base2-gpt-5-worker agent inline after you have gathered all the context you need (and not before!).',
+    isNormal && '- Only make edits after generating a plan.',
+    !isFast &&
+      '- Code reviewers should be spawned after you have made your edits.',
+    !isFast &&
+      '- Validators should be spawned after you have done a code review.',
+  ).join('\n  ')}
 - **No need to include context:** When prompting an agent, realize that many agents can already see the entire conversation history, so you can be brief in prompting them without needing to include context.
-- **Don't spawn code reviewers${!isFast ? '/validators' : ''} for trivial changes or quick follow-ups:** You should spawn the code reviewer${!isFast ? '/validator' : ''} for most changes, but not for little changes or simple follow-ups.
+${!isFast ? "- **Don't spawn code reviewers/validators for trivial changes or quick follow-ups:** You should spawn the code reviewer/validator for most changes, but not for little changes or simple follow-ups." : ''} 
 
 # Core Mandates
 
@@ -132,21 +142,23 @@ The user asks you to implement a new feature. You respond in multiple steps:
 2a. Read all the relevant files using the read_files tool.
 ${
   isFast
-    ? `3. Write out your implementation plan as a bullet point list.`
-    : '3. Spawn a generate-plan agent to generate a plan for the changes.'
-}
+    ? `3. Write out your implementation plan as a bullet point list.
+4. Use the str_replace or write_file tool to make the changes.
+5. Inform the user that you have completed the task in one sentence without a final summary. Don't create any markdown summary files either, unless asked by the user.`
+    : isNormal
+      ? `3. Spawn a generate-plan agent to generate a plan for the changes.
 4. Use the str_replace or write_file tool to make the changes.
 5. Spawn a code-reviewer to review the changes. Consider making changes suggested by the code-reviewer.
-${
-  isFast
-    ? `6. Inform the user that you have completed the task in one sentence without a final summary.`
-    : `6. Spawn a validator to run validation commands (tests, typechecks, etc.) to ensure the changes are correct.
+6. Spawn a validator to run validation commands (tests, typechecks, etc.) to ensure the changes are correct.
 7. Inform the user that you have completed the task in one sentence without a final summary. Don't create any markdown summary files either, unless asked by the user.`
+      : `3. IMPORTANT: You must spawn a base2-gpt-5-worker agent inline (with spawn_agent_inline tool) to do the planning and editing.
+4. Fix any issues left by the base2-gpt-5-worker agent.
+5. Inform the user that you have completed the task in one sentence without a final summary. Don't create any markdown summary files either, unless asked by the user.`
 }`,
 
-    stepPrompt: `Don't forget to spawn agents that could help, especially: the file-picker-max and find-all-referencer to get codebase context,${!isFast ? ' the generate-plan agent to create a plan,' : ''} code-reviewer to review changes${!isFast ? ', and the validator to run validation commands' : ''}.`,
+    stepPrompt: `Don't forget to spawn agents that could help, especially: the file-picker-max and code-searcher to get codebase context,${!isFast ? ' the generate-plan agent to create a plan,' : ''} code-reviewer to review changes${!isFast ? ', and the validator to run validation commands' : ''}.`,
 
-    handleSteps: function* ({ prompt, params }) {
+    handleSteps: function* ({ params }) {
       let steps = 0
       while (true) {
         steps++

.agents/planners/planner-pro.ts

@@ -9,8 +9,7 @@ const definition: SecretAgentDefinition = {
   model: 'openai/gpt-5-pro',
   publisher,
   displayName: 'Planner Pro',
-  spawnerPrompt:
-    'Generates 5 alternative plans for a user request, analyzes them, and selects the best and simplest one that meets all requirements.',
+  spawnerPrompt: 'Uses deep thinking to generate an implementation plan for a user request.',
   inputSchema: {},
   outputMode: 'last_message',
   spawnableAgents: [],

evals/buffbench/main.ts

@@ -5,8 +5,8 @@ import { runBuffBench } from './run-buffbench'
 async function main() {
   await runBuffBench({
     evalDataPath: path.join(__dirname, 'eval-codebuff.json'),
-    agents: ['research-implement-orchestrator', 'base'],
-    taskConcurrency: 20,
+    agents: ['base2-fast', 'base2', 'base-max'],
+    taskConcurrency: 10,
   })
 
   process.exit(0)