class LittleGhost::Tools::WriteTodos

WriteTodos lets an agent share a live plan with the application and the person following its work. Add it like any other tool:

class ResearchAgent < LittleGhost::Agent
  tools LittleGhost::Tools::WriteTodos
end

Each execution replaces the full plan in RunContext state. Todo IDs remain stable across updates, statuses are pending, in_progress, or completed, and no more than one todo may be in progress. When no context is available, the tool retains fallback state on its instance.

Public Instance Methods

call (input)

Source

# File lib/little_ghost/tools/write_todos.rb, line 53
def call(input)
  todos = input.fetch("todos")
  if todos.count { |todo| todo["status"] == "in_progress" } > 1
    raise ToolError, framework_prompt("tools/built_in/write_todos/feedback/single_progress")
  end

  ids = todos.map { |todo| todo.fetch("id") }
  raise ToolError, framework_prompt("tools/built_in/write_todos/feedback/unique_ids") unless ids.uniq.length == ids.length

  if context
    context.state["little_ghost.plan"] ||= empty_plan
    state = context.state.fetch("little_ghost.plan")
  else
    state = @fallback_state ||= empty_plan
  end
  state.replace("plan_title" => input.fetch("plan_title"), "todos" => todos.map(&:dup))
  state.dup
end

Replaces the stored plan after enforcing progress and ID invariants.

description ()

Source

# File lib/little_ghost/tools/write_todos.rb, line 48
def description
  framework_prompt("tools/built_in/write_todos/description")
end

execute (input, context: nil)

Source

# File lib/little_ghost/tools/write_todos.rb, line 44
def execute(input, context: nil)
  super(normalize_titles(input), context:)
end

Trims user-facing titles before normal tool validation and execution.

Calls superclass method LittleGhost::Tool#execute