---
title: Concrete Actions Only
description: The single highest-leverage convention in the system — every task title names an action you can do or a decision you can make — and the companion rule that new initiatives always get a note and a task, never a bare "scope it".
track: core
url: https://www.chiefofstaff.io/guides/concrete-actions-only
published: 2026-09-01
author: chiefofstaff.io editorial system
outcome: rewrite any task titled "define / scope / plan / look into X" into its actual first action or decision, and refuse to create a task for a new initiative without also creating its note.
prerequisites: the-task-graph
source: chiefofstaff.io
---

# Concrete Actions Only

The single highest-leverage convention in the system — every task title names an action you can do or a decision you can make — and the companion rule that new initiatives always get a note and a task, never a bare "scope it".

## The Rule

> Every task title names a concrete next action the owner can *do*, or a specific decision they can *make*. Banned as titles: "define / scope / figure out / think about / plan / look into X."

That is the whole rule. It came from a single correction. The assistant had captured a new internal initiative and a person who might own it, created the project note and the person note — all correct — and then titled the task "Define the owner's task scope." The human's reaction, paraphrased: *what does that mean for me? What am I supposed to do with that on a Tuesday morning?*

A task is a **commitment device**. "Scope X" commits to nothing. It sits in the list, ages, and generates a small guilt each time it is skimmed, because there is no moment at which it becomes obviously do-able. It is abstraction wearing a task's clothes.

## What to Do Instead

When the assistant catches itself writing an abstraction, it decomposes: *what is the actual first action?*

- "Scope the new reporting tool" → "Write the five questions the reporting tool must answer; send to M for a yes/no on each."
- "Think about the pricing change" → "Decide: raise the base tier to the new price on the 1st, or hold until the contract renews — needs last quarter's churn number from finance."
- "Look into the visa situation" → "Call the immigration lawyer; ask for the timeline and the document list."

If the inputs genuinely do not exist yet, the task's job is to **name the missing inputs**: "Decide X — needs A from the accountant and B from you." That is still a commitment: to obtain A and B. It is not deferring the thinking; it is stating precisely what the thinking is waiting on.

## The Companion Rule: A Note and a Task, Never a Bare Task

When a capture introduces a new initiative, project, venture, or person, the assistant produces **both**:

1. The durable note — a project note under the right venture, or a person note in `Atlas/people/` — with provenance.
2. A concrete-action task.

Never a bare task with no note behind it. A task titled "Follow up with new contact about the tool idea" that references a person and a project which exist nowhere in the vault is a task that will make no sense in three weeks. The note is where the context lives; the task is the next move. Curators and the inbox triager cannot write to `Atlas/`, so when they draft such a task they flag "needs a note" and the chief of staff creates it before the task is confirmed.

## Why It Is the Highest-Leverage Rule

Three reasons, in increasing order of importance:

- **It makes the brief actionable.** A brief line that says "scope X" asks nothing. A line that says "Decide: A or B — needs C" can be answered with one word from a phone.
- **It exposes missing inputs early.** The moment you try to write the first action, you discover what you don't know. That discovery is the work; the abstraction was hiding it.
- **It keeps the task graph honest.** A backlog full of "look into" items looks like a plan and is actually a list of anxieties. A backlog of concrete actions can be walked in a weekly review and each item either done, delegated, deferred with a date, or deleted.

## Framing Brief Lines the Same Way

The rule extends from task titles to the brief. Each input item is a complete sentence framed as the actual decision — "Subscription renews Friday (3-day window). Keep or cancel?" — not a topic plus a task ID. Task IDs become small reference tags at the end of the line, so you can act on them from the review tool, but you never type one. Voice replies work by position ("for the subscription one, cancel; for the invite, decline"), which only works if every line is phrased as something with an obvious answer.

## The Test

Read the title and ask: *could I start this in the next ten minutes if the calendar were empty?* If the honest answer is "I'd have to think about what it means first," it is not a task yet.