Topic Focused Workflow (TFW) now has two published forms: a reference site at tfw.how, and an agent skill that applies the method inside AI tools. This post explains why they exist, what each one is for, how you can use them, and how AspirationalX uses the method internally.
If you want the idea itself first, start with the earlier essay: Topic Focused Workflow: A Better Way to Organize Commitments and Responsibilities. The short version is one rule:
Define the topic first, then make the tools reflect that topic consistently.
Why an essay was not enough
The essay explains why the method matters. It does not give anyone enough detail to apply it the same way twice.
That gap showed up in two places.
People. When someone set up a new space, project, or channel, they needed exact answers. What counts as a topic? How should a surface be named? Where does a decision go, and where does the discussion about it go? An essay answers those questions in general. A method has to answer them precisely enough that two people working separately reach the same result.
AI agents. More of the filing, routing, and setup work in my tools is now done by agents. An agent follows the definitions it is given. If a term is loose, the agent fills the gap with a guess, and the guess becomes a new folder, a new channel, or a new code that does not match anything else. That is the same drift TFW exists to prevent, only faster.
So the method needed two things: a stable reference with exact definitions, and a way for an agent to apply those definitions and stop when the answer is not clear.
What tfw.how is
tfw.how is the reference for the method. It is written for two readers at once: people who set up or maintain tools for themselves, a team, or an organization, and AI agents that read, file, or route work on behalf of those people.
The structure of the site follows the structure of the method:
- Method: six steps. Choose the topic. Use the most specific valid topic. Align tool naming across surfaces. Keep hierarchies shallow. Separate conversation from the source of truth. Define participants, access, and maintainers.
- Model: the definitions underneath the steps. A topic is exactly one venture combined with exactly one unit, such as one company’s Operations. Each topic gets a four-character code: two characters for the venture and two for the unit. Every surface for that topic carries the code, so the code also marks the access boundary.
- Focus System: how a topic records why its work matters. Objectives, intentions, strategies, and actions are linked with two phrases only: an action “contributes to” a strategy, and the strategy is “fulfilled by” the action.
- Tool classes and tools: what each class of tool is for (task management, knowledge base, document storage, chat, email, calendar, code repository, AI agent workspace), and how the method maps onto specific products such as Jira, Confluence, Slack, and Google Drive.
- Glossary: every term, defined once, so a person and an agent read the same word the same way.
All examples on the site use an invented company, Example Co, with the topic code EXOP for its Operations. That keeps the method separate from any one organization’s setup.
Built to be read by AI
The site publishes its content in forms an agent can use directly:
/llms.txtlists every page with a one-sentence description, so an agent can choose what to read./llms-full.txtholds the source Markdown of every page in one file, in navigation order.- Every page has a Markdown version at its URL plus
index.md, for examplehttps://tfw.how/model/index.md. - Each page carries structured data, and the glossary is published as a defined term set.
What the skill is
The second piece is an agent skill called Topic Context Governance. A skill is a packaged set of instructions an AI tool loads when a task calls for it. This one applies TFW to the work in front of the agent. It can:
- set up a topic: find the registry, take the code from it, name each required surface, and create the Focus System records
- answer “where does this belong?” with the topic, the kind of information, the source of truth, and which other surfaces should link to it
- audit an existing setup and report each finding with the evidence it searched and the smallest repair
The most important part of the skill is what it will not do. It takes codes from the user’s own registry and never invents one. When no topic clearly matches, it reports a registry gap and asks, instead of creating a parallel space. It proposes changes and makes none without approval.
That restraint is the reason the skill exists. An agent that creates structure on its own will create plausible structure that does not match the rest of the system. An agent that stops at a gap and asks keeps the system coherent.
The skill includes conventions for Jira, Confluence, Google Drive, Slack, Gmail, Google Calendar, GitHub, ChatGPT, and Claude, and it is packaged for Claude and for ChatGPT and Codex.
How you can use it
You can use the method today through tfw.how:
- Read it. Start with the Method page, then the Model.
- Give it to your assistant. Paste
https://tfw.how/llms-full.txtinto Claude, ChatGPT, or another assistant and ask it to work from that file. For one subject, give it that page’sindex.mdinstead. - Use the prompts. The Use with AI page has four copyable prompts: set up a topic, audit a tool, name a new surface, and decide where something belongs. Each one tells the assistant to use the glossary’s terms exactly and to propose changes, not make them.
The skill is not public yet. AspirationalX uses it internally, and a public version will be published after review. The Use with AI page on tfw.how will link to it when it is available.
The content of tfw.how is currently all rights reserved. You can read it and use it with your own assistant. It is not yet released under an open license.
How AspirationalX uses it
AspirationalX runs on TFW. Every durable context in the organization is a topic, one venture combined with one unit, and each topic has its own code.
In practice, each topic has:
- one task management project, which is also the registry entry for the topic
- one knowledge base space, with a root page that states what the space is for and what it is not for
- one shared drive
- exactly two chat channels: one for people, and one for output from connected software
- calendar events titled with the topic code
Every task links upward to the strategy it serves. Strategies link to intentions, and intentions to objectives. With those links in place, a piece of work shows which topic it belongs to, why it exists, and where its records live.
This website is an example. Each working session on it has one ticket in the website’s topic. That ticket links to the standing website strategy, and every commit is marked with the ticket’s key. The ticket for this post links to the website strategy and to the strategy for telling others what I am working on.
The skill is how agents work inside that structure. When an agent files a record, names a new surface, or checks where something belongs, it reads the live registry and follows the same rules a person would. Conventions that are specific to AspirationalX, such as particular tools and the real registry, are kept out of the public material. The public site and the skill describe the method, not one organization’s configuration of it.
What this changes
Before tfw.how, applying TFW outside AspirationalX depended on reading my essay and asking me what I meant. Now the definitions are written down once, in a form people and agents both read, and the skill applies them with the same limits a careful person would use.
If you keep work across several tools, try one of the prompts on Use with AI against a single topic you already have. The audit prompt will show you where your tool names, boundaries, and sources of truth have drifted apart.