Combine agent-builder and base-agent builder. Add diff-reviewer1, 2, and 3 as example agents.

Author: jahooma

.agents/agent-builder.ts

@@ -0,0 +1,18 @@
+import { AgentConfig } from '@codebuff/common/util/types/agent-config.d'
+
+const config: AgentConfig = {
+  id: 'diff-reviewer-1',
+
+  displayName: 'Diff Reviewer (Level 1)',
+  model: 'openai/gpt-5',
+  toolNames: ['read_files', 'run_terminal_command'],
+
+  parentPrompt: 'Spawn when you need to review code changes in the git diff',
+
+  instructionsPrompt: `Execute the following steps:
+1. Run git diff
+2. Read the files that have changed
+3. Review the changes and suggest improvements`,
+}
+
+export default config

.agents/examples/diff-reviewer-1.ts

@@ -0,0 +1,54 @@
+import {
+  AgentConfig,
+  AgentStepContext,
+} from '@codebuff/common/util/types/agent-config.d'
+
+const config: AgentConfig = {
+  id: 'diff-reviewer-2',
+  displayName: 'Diff Reviewer (Level 2)',
+  model: 'openai/gpt-5',
+
+  inputSchema: {
+    prompt: {
+      type: 'string',
+      description:
+        'Please provide a short description of the changes you want to review',
+    },
+  },
+  toolNames: ['read_files', 'run_terminal_command'],
+
+  parentPrompt: 'Spawn when you need to review code changes in the git diff',
+
+  systemPrompt:
+    'You are an expert software developer. Your job is to review code changes and provide helpful feedback.',
+
+  instructionsPrompt: `Execute the following steps:
+1. Run git diff
+2. Read the files that have changed
+3. Review the changes and suggest improvements
+
+Use the following guidelines while reviewing the changes:
+- Find ways to simplify the code
+- Reuse existing code as much as possible instead of writing new code
+- Preserve as much behavior as possible in the existing code
+- Prefer changing as few lines of code as possible
+- Look for opportunities to improve the code's readability
+- Look for logical errors in the code
+- Look for missed cases in the code
+- Look for any other bugs`,
+
+  handleSteps: function* ({ agentState, prompt, params }: AgentStepContext) {
+    // Step 1: Run git diff immediately. Saves the agent a step, lowering cost and latency!
+    yield {
+      toolName: 'run_terminal_command',
+      args: {
+        command: 'git diff',
+      },
+    }
+
+    // Step 2: Let AI run the rest of the steps!
+    yield 'STEP_ALL'
+  },
+}
+
+export default config

.agents/examples/diff-reviewer-2.ts

@@ -0,0 +1,99 @@
+import {
+  AgentConfig,
+  AgentStepContext,
+} from '@codebuff/common/util/types/agent-config.d'
+
+const config: AgentConfig = {
+  id: 'diff-reviewer-3',
+
+  displayName: 'Diff Reviewer (Level 3)',
+  model: 'openai/gpt-5',
+  inputSchema: {
+    prompt: {
+      type: 'string',
+      description:
+        'Please provide a short description of the changes you want to review',
+    },
+  },
+  outputMode: 'last_message',
+
+  toolNames: ['read_files', 'run_terminal_command', 'spawn_agents'],
+  subagents: ['james/file-explorer@0.1.3'],
+
+  parentPrompt: 'Spawn when you need to review code changes in the git diff',
+
+  systemPrompt:
+    'You are an expert software developer. Your job is to review code changes and provide helpful feedback.',
+
+  instructionsPrompt: `Review the changes and suggest improvements.
+
+Use the following guidelines while reviewing the changes:
+- Find ways to simplify the code
+- Reuse existing code as much as possible instead of writing new code
+- Preserve as much behavior as possible in the existing code
+- Prefer changing as few lines of code as possible
+- Look for opportunities to improve the code's readability
+- Look for logical errors in the code
+- Look for missed cases in the code
+- Look for any other bugs`,
+
+  handleSteps: function* ({ agentState, prompt, params }: AgentStepContext) {
+    // Step 1: Get list of changed files from git diff --name-only
+    const { toolResult: gitDiffFilesResult } = yield {
+      toolName: 'run_terminal_command',
+      args: {
+        command: 'git diff --name-only',
+      },
+    }
+
+    // Then, extract file paths from the result
+    const changedFiles = (gitDiffFilesResult || '')
+      .split('\n')
+      .map((line) => line.trim())
+      .filter((line) => line && !line.startsWith('??') && !line.includes('OSC'))
+
+    // Step 2: Read the files
+    if (changedFiles.length > 0) {
+      yield {
+        toolName: 'read_files',
+        args: {
+          paths: changedFiles,
+        },
+      }
+    }
+
+    // Step 3: Run full git diff to see the actual changes
+    yield {
+      toolName: 'run_terminal_command',
+      args: {
+        command: 'git diff',
+      },
+    }
+
+    // Step 4: Put words in the AI's mouth to get it to spawn the file explorer.
+    yield {
+      toolName: 'add_message',
+      args: {
+        role: 'assistant',
+        content:
+          'Now I will spawn a file explorer to find any missing codebase context.',
+      },
+    }
+
+    yield 'STEP'
+
+    // Step 5: Put words in the AI's mouth to review the changes.
+    yield {
+      toolName: 'add_message',
+      args: {
+        role: 'assistant',
+        content: 'Here is my comprehensive review of the changes.',
+      },
+    }
+
+    // Step 6: Let AI review the changes in a final step. (The last message is also the agent's output.)
+    yield 'STEP'
+  },
+}
+
+export default config

.agents/examples/diff-reviewer-3.ts

@@ -10,7 +10,7 @@ const config: AgentConfig = {
   parentPrompt:
     'Spawns multiple file picker agents in parallel to comprehensively explore the codebase from different perspectives',
   model: 'anthropic/claude-4-sonnet-20250522',
-  outputMode: 'json',
+  outputMode: 'structured_output',
   includeMessageHistory: false,
   toolNames: ['spawn_agents', 'set_output'],
   subagents: [`file-picker`],