Skip to main content

Events Reference

Extensions can listen to and modify events throughout AiderDesk. This page documents all available events and their properties.

Modifying Event Properties​

Properties in event interfaces can be modified by returning a partial event from your handler. The rule is simple:

  • Properties marked readonly cannot be modified
  • Properties without readonly can be modified by returning them in a partial event
// Example: Modifying a modifiable property
async onAgentStarted(event: AgentStartedEvent, context: ExtensionContext) {
  // prompt is NOT readonly, so it can be modified
  return {
    prompt: event.prompt?.toUpperCase(),
  };
}

Event Handling Patterns​

1. Read-Only​

Observe the event without modifying anything.

async onTaskClosed(event: TaskClosedEvent, context: ExtensionContext) {
  context.log(`Task ${event.task.name} closed`, 'info');
  // Return nothing - event continues unchanged
}

2. Blocking​

Prevent an operation by setting blocked: true.

async onToolApproval(event: ToolApprovalEvent, context: ExtensionContext) {
  if (event.toolName === 'dangerous_tool') {
    return { blocked: true };  // Prevents execution
  }
}

3. Modifying​

Change event data by returning a partial event.

async onAgentStarted(event: AgentStartedEvent, context: ExtensionContext) {
  return {
    agentProfile: {
      ...event.agentProfile,
      customInstructions: event.agentProfile.customInstructions + '\nBe concise.',
    },
  };
}

Event Categories​

CategoryEventsPurpose
TaskCreated, Prepared, Initialized, Closed, UpdatedTask lifecycle
ProjectStarted, StoppedProject lifecycle
AgentStarted, Step Started, Finished, Step FinishedAgent execution
Message OptimizationOptimize, Important RemindersMessage optimization
ToolApproval, Called, FinishedTool execution
FileAdded, Dropped, Rule Files RetrievedFile context management
PromptStarted, FinishedPrompt processing
Prompt TemplateTemplate RenderedPrompt template customization
ResponseChunk, CompletedResponse streaming
ApprovalHandle ApprovalUser approvals
SubagentStarted, FinishedSubagent execution
QuestionAsked, AnsweredUser questions
CommandExecuted, Custom Command ExecutedCommand execution
CommitBefore Commit, After CommitCommit lifecycle
AiderPrompt Started, Prompt FinishedLegacy Aider events

Task Events​

TaskCreatedEvent​

Called when a new task is created.

interface TaskCreatedEvent {
  task: TaskData;
}

TaskPreparedEvent​

Called when a task is prepared (both new and loaded tasks).

interface TaskPreparedEvent {
  task: TaskData;
}

TaskInitializedEvent​

Called when a task is initialized and ready for use.

interface TaskInitializedEvent {
  readonly task: TaskData;
}

TaskClosedEvent​

Called when a task is closed.

interface TaskClosedEvent {
  readonly task: TaskData;
}

TaskUpdatedEvent​

Called before a task is updated and saved.

interface TaskUpdatedEvent {
  task: TaskData;
}

Project Events​

ProjectStartedEvent​

Called when a project is started.

interface ProjectStartedEvent {
  readonly baseDir: string;
}

ProjectStoppedEvent​

Called when a project is stopped.

interface ProjectStoppedEvent {
  readonly baseDir: string;
}

Prompt Events​

PromptStartedEvent​

Called when prompt processing starts.

interface PromptStartedEvent {
  prompt: string;
  mode: Mode;
  promptContext: PromptContext;
  blocked?: boolean;
}

PromptFinishedEvent​

Called when prompt processing finishes.

interface PromptFinishedEvent {
  responses: ResponseCompletedData[];
}

Prompt Template Events​

PromptTemplateEvent​

Called when a prompt template is rendered. Use to customize or override prompt templates.

interface PromptTemplateEvent {
  /** Template name (e.g., 'system-prompt', 'init-project', etc.) */
  readonly name: string;
  /** Template data object */
  readonly data: unknown;
  /** Rendered prompt that can be overridden by extension */
  prompt: string;
}

Available Template Names:

  • system-prompt - System prompt for the agent
  • init-project - Project initialization prompt
  • workflow - Workflow execution prompt
  • compact-conversation - Conversation compaction prompt
  • commit-message - Git commit message generation prompt
  • task-name - Task name generation prompt
  • conflict-resolution - Conflict resolution prompt
  • conflict-resolution-system - Conflict resolution system prompt
  • update-task-state - Task state update prompt
  • handoff - Task handoff prompt
  • code-inline-request - Inline code request prompt

Example: Customizing System Prompt​

async onPromptTemplate(event: PromptTemplateEvent, context: ExtensionContext) {
  // Customize the system prompt
  if (event.name === 'system-prompt') {
    return {
      prompt: event.prompt + '\n\nAdditional instructions: Always be concise.',
    };
  }
}

Example: Project-Specific Prompt Customization​

async onPromptTemplate(event: PromptTemplateEvent, context: ExtensionContext) {
  const projectDir = context.getProjectDir();

  // Add project-specific context to init-project prompt
  if (event.name === 'init-project') {
    const customInstructions = `

## Project-Specific Guidelines
This project uses TypeScript with strict mode enabled.
Always prefer type-safe implementations over any types.
    `;

    return {
      prompt: event.prompt + customInstructions,
    };
  }
}

Agent Events​

AgentStartedEvent​

Called when agent mode starts. Use to modify prompts, context, or block execution.

interface AgentStartedEvent {
  readonly mode: Mode;
  prompt: string | null;
  agentProfile: AgentProfile;
  providerProfile: ProviderProfile;
  model: string;
  promptContext?: PromptContext;
  systemPrompt: string | undefined;
  contextMessages: ContextMessage[];
  contextFiles: ContextFile[];
  blocked?: boolean;
  images?: string[];
  skillsToActivate?: string[];
  modelCallSettings?: ModelCallSettings;
}

modelCallSettings contains the computed model-call defaults when the event is dispatched. Return a partial value to override AI SDK call parameters; it is merged with the defaults. See Model Call Settings for the available fields.

AgentStepStartedEvent​

Called before each agent step starts (before the LLM call). Use to modify messages that will be sent.

interface AgentStepStartedEvent {
  readonly mode: Mode;
  readonly agentProfile: AgentProfile;
  readonly currentResponseId: string;
  readonly iterationCount: number;
  messages: ContextMessage[];
}

AgentFinishedEvent​

Called when agent mode finishes.

interface AgentFinishedEvent {
  readonly mode: Mode;
  readonly aborted: boolean;
  readonly contextMessages: ContextMessage[];
  resultMessages: ContextMessage[];
}

AgentStepFinishedEvent​

Called after each agent step completes.

interface AgentStepFinishedEvent {
  readonly mode: Mode;
  readonly agentProfile: AgentProfile;
  readonly currentResponseId: string;
  readonly stepResult: AgentStepResult;
  finishReason: 'stop' | 'length' | 'content-filter' | 'tool-calls' | 'error' | 'other' | 'unknown';
  responseMessages: ContextMessage[];
}

Message Optimization Events​

OptimizeMessagesEvent​

Called when messages are being optimized. Use to modify or filter the optimized messages.

interface OptimizeMessagesEvent {
  readonly originalMessages: ContextMessage[];
  optimizedMessages: ContextMessage[];
}

ImportantRemindersEvent​

Called when important reminders are being generated. Use to modify the reminders content.

interface ImportantRemindersEvent {
  readonly agentProfile: AgentProfile;
  remindersContent: string;
}

Tool Events​

ToolApprovalEvent​

Called when a tool requires approval.

interface ToolApprovalEvent {
  readonly toolName: string;
  readonly input: Record<string, unknown> | undefined;
  blocked?: boolean | string;
  allowed?: boolean;
}

ToolCalledEvent​

Called when a tool is about to be executed.

interface ToolCalledEvent {
  readonly toolName: string;
  readonly agentProfile: AgentProfile;
  readonly abortSignal?: AbortSignal;
  input: Record<string, unknown> | undefined;
  output?: unknown;
}

ToolFinishedEvent​

Called after tool execution completes.

interface ToolFinishedEvent {
  readonly toolName: string;
  readonly agentProfile: AgentProfile;
  readonly input: Record<string, unknown> | undefined;
  output: unknown;
}

File Events​

FilesAddedEvent​

Called when files are added to context. Return empty array to prevent addition.

interface FilesAddedEvent {
  files: ContextFile[];
}

FilesDroppedEvent​

Called when files are dropped into the chat. Return empty array to prevent addition.

interface FilesDroppedEvent {
  files: ContextFile[];
}

RuleFilesRetrievedEvent​

Called when rule files are retrieved. Modify to filter or add rule files.

interface RuleFilesRetrievedEvent {
  files: ContextFile[];
}

Response Events​

ResponseChunkEvent​

Called on each response chunk during streaming.

interface ResponseChunkEvent {
  chunk: ResponseChunkData;
}

ResponseCompletedEvent​

Called when response is complete.

interface ResponseCompletedEvent {
  response: ResponseCompletedData;
}

Approval Events​

HandleApprovalEvent​

Called when handling user approval requests.

interface HandleApprovalEvent {
  key: string;
  text: string;
  subject?: string;
  blocked?: boolean;
  allowed?: boolean;
}

Subagent Events​

SubagentStartedEvent​

Called when a subagent starts. Use to modify or block subagent execution.

interface SubagentStartedEvent {
  subagentProfile: AgentProfile;
  prompt: string;
  promptContext?: PromptContext;
  contextMessages: ContextMessage[];
  contextFiles: ContextFile[];
  systemPrompt?: string;
  blocked?: boolean;
}

SubagentFinishedEvent​

Called when a subagent finishes.

interface SubagentFinishedEvent {
  readonly subagentProfile: AgentProfile;
  resultMessages: ContextMessage[];
}

Question Events​

QuestionAskedEvent​

Called when a question is asked to the user. Set answer to auto-answer.

interface QuestionAskedEvent {
  question: QuestionData;
  answer?: string;
}

QuestionAnsweredEvent​

Called when the user answers a question.

interface QuestionAnsweredEvent {
  readonly question: QuestionData;
  answer: string;
  userInput?: string;
}

Command Events​

CommandExecutedEvent​

Called when a slash command is executed.

interface CommandExecutedEvent {
  command: string;
  blocked?: boolean;
}

CustomCommandExecutedEvent​

Called when a custom command is executed.

interface CustomCommandExecutedEvent {
  command: CustomCommand;
  mode: Mode;
  blocked?: boolean;
  prompt?: string;
}

Commit Events​

BeforeCommitEvent​

Called before changes are committed. Use to modify the commit message, amend flag, or block the commit.

interface BeforeCommitEvent {
  message: string;
  amend: boolean;
  blocked?: boolean;
}

AfterCommitEvent​

Called after changes are committed. Read-only event for observing committed changes.

interface AfterCommitEvent {
  readonly message: string;
  readonly amend: boolean;
}

Aider Events (Legacy)​

AiderPromptStartedEvent​

Called when Aider prompt starts (legacy event).

interface AiderPromptStartedEvent {
  prompt: string;
  mode: Mode;
  promptContext: PromptContext;
  messages: ConnectorMessage[];
  files: ContextFile[];
  blocked?: boolean;
  autonomyMode?: 'manual' | 'guided' | 'autonomous';
  denyCommands?: boolean;
}

AiderPromptFinishedEvent​

Called when Aider prompt finishes (legacy event).

interface AiderPromptFinishedEvent {
  responses: ResponseCompletedData[];
}