Todo & task tracking

On long, multi-step tasks it helps the model (and you) to track todos explicitly. A dedicated todo overlay keeps the plan visible and — crucially — survives /reload and context compaction, so the agent doesn’t lose the thread on big jobs.

Alternatives

1. @juicesharp/rpiv-todo

A todo list for the model, rendered as a live overlay that persists through /reload and conversation compaction.

pi install npm:@juicesharp/rpiv-todo

Benefits

• Persistent: the list isn’t lost when older messages are summarized.
• Live overlay keeps progress visible in the TUI.
• Part of the cohesive rpiv-* family.

Drawbacks

• One more UI element on screen.
• Most valuable on long tasks; light tasks don’t need it.

2. Sub-agent / orchestration packages with built-in task lists

Several orchestration tools track tasks as part of their workflow. For example, @pi-agents/orchid manages batch tasks and loops, and sub-agent extensions track delegated work items.

pi install npm:@pi-agents/orchid

Benefits

• Task tracking integrated with execution (delegated/parallel work).
• No separate todo tool to wire up.

Drawbacks

• Tied to that tool’s workflow; not a standalone todo overlay.
• Heavier if all you want is a checklist.

3. A prompt template + AGENTS.md convention (no package)

The zero-dependency option: add a prompt template that asks the model to maintain a checklist, and document the convention in AGENTS.md.

~/.pi/agent/prompts/todo.md

Maintain a markdown checklist of steps for this task. Update it after each step.

Benefits

• No third-party code; fully transparent.
• Trivial to customize.

Drawbacks

• Not persistent through compaction unless you re-state it.
• No dedicated UI.

Which should I pick?

• Want a real, persistent todo overlay: @juicesharp/rpiv-todo.
• Already orchestrating multi-agent work: use that tool’s built-in tracking.
• Want zero dependencies: a prompt template + AGENTS.md convention.