Problem

Currently, hooks receive different context depending on which hook type they are:

• MESSAGE_POST_PROCESS: Receives log, workspace, manager
• TOOL_POST_EXECUTE: Only receives tool_name and tool_use

This inconsistency makes it difficult to implement awareness tools that need access to conversation state.

Current Implementation

# In ToolUse.execute()
def execute(...):
    # ...
    result = tool_func(...)
    run_hook(
        hook_name="TOOL_POST_EXECUTE",
        tool_name=tool.name,
        tool_use=self,
    )

Desired Implementation

All hooks should receive the same baseline context:

def execute(self, log, workspace, manager):
    # ...
    result = tool_func(...)
    run_hook(
        hook_name="TOOL_POST_EXECUTE",
        log=log,
        workspace=workspace,
        manager=manager,
        tool_name=tool.name,
        tool_use=self,
    )

Use Case

The token-awareness and time-awareness tools need to show budget/time information after each tool execution (similar to how Anthropic shows token usage after tool results). Currently, they can only use MESSAGE_POST_PROCESS because TOOL_POST_EXECUTE doesn't have access to log (needed for token calculation) or workspace (needed for tracking state).

Impact

• Makes hook system more consistent and predictable
• Enables awareness tools to provide feedback after individual tool executions
• Reduces coupling between hook types and their capabilities

Related

• PR feat: add token & time awareness tools/hooks #665: Token and time awareness tools (currently using MESSAGE_POST_PROCESS as workaround)