Use with AI

TFW is meant to be applied by people and by AI agents in the same way. This page lists the machine-readable versions of this site and gives prompts that you can copy into an AI assistant.

Machine-readable versions of this site

Resource	Address	What it contains
llms.txt	/llms.txt	An index of every page, grouped by section, with a one-sentence description of each page. It follows the llms.txt format.
llms-full.txt	/llms-full.txt	The source Markdown of every page in one plain-text file, in navigation order.
Markdown per page	The page address followed by `index.md`	The source Markdown of one page, with its title, address, and description at the top.
JSON-LD	In the HTML of every page	Structured data from schema.org: `WebSite` on the home page, `TechArticle` on every other page, `BreadcrumbList` on nested pages, and `DefinedTermSet` on the glossary.
Sitemap	/sitemap.xml	The address of every page.

The Markdown address of a page is its address followed by index.md. For example:

Page	Markdown version
`https://tfw.how/`	`https://tfw.how/index.md`
`https://tfw.how/model/`	`https://tfw.how/model/index.md`
`https://tfw.how/tool-classes/chat/`	`https://tfw.how/tool-classes/chat/index.md`
`https://tfw.how/tools/slack/`	`https://tfw.how/tools/slack/index.md`

Every HTML page also links to its Markdown version with <link rel="alternate" type="text/markdown">, and the footer of every page has a link named “This page as Markdown”.

Which version to use:

To give an assistant the whole method in one step, use llms-full.txt. It is the most complete source.
To give an assistant one subject, such as the guidance for one tool, use the index.md of that page.
To let an agent decide which pages to read, give it llms.txt.

Prompts

Each prompt below is written for a general-purpose AI assistant. Copy a prompt, replace the text in square brackets, and paste it into the assistant. If the assistant cannot open web addresses, download llms-full.txt and attach it instead.

The prompts ask the assistant to propose changes, not to make them. Review the proposal before you or the assistant change any tool.

Set up a topic

Read https://tfw.how/llms-full.txt. It describes Topic Focused Workflow (TFW).
Use its terms exactly as defined in its glossary.

Help me set up one topic.
Venture: [name of the business, organization, household, or person]
Unit: [functional area, for example Operations or Finance]
Tools we use: [for example Google Drive, Jira, Confluence, Slack, Gmail, Google Calendar]
Existing topic codes: [list them, or write "none"]

Produce a topic profile with:
1. The canonical topic name and a proposed four-character topic code
   (two characters for the venture, two for the unit). Check that the code
   does not repeat an existing code.
2. One surface per tool class, with its exact name, following the naming
   rules on the tool pages. Chat has two channels: chat and automated.
3. The owner, participants, access boundary (one group per topic),
   and a maintainer for each surface. Mark anything I have not told you as unknown.
4. The lifecycle status of the topic.
5. A list of questions I need to answer before the setup is complete.

Do not invent names of people. Do not create anything; only propose.

Audit a tool against TFW

Read https://tfw.how/llms-full.txt. It describes Topic Focused Workflow (TFW).
Use its terms exactly as defined in its glossary.

Audit the following [tool name] structure against TFW.
Our topics and topic codes: [list each topic code and canonical name]

[Paste a list of the shared drives, projects, spaces, channels, groups,
or calendars in the tool, with their names, descriptions, and who has access.]

For each item, report:
1. The topic it maps to, or "no match" (a registry gap).
2. Whether the name follows the TFW naming rule for this tool class.
3. Whether it duplicates another surface of the same topic (a provisioning gap).
4. Whether its access matches the topic's access boundary.
5. Whether it has a description that states its purpose and links to the
   source of truth.

Then list topics that have no surface in this tool, and give the smallest
change that fixes each problem. Do not assume a mapping from a similar name;
mark uncertain mappings as uncertain.

Name a new surface

Read https://tfw.how/llms-full.txt. It describes Topic Focused Workflow (TFW).
Use its terms exactly as defined in its glossary.

I need a new surface for one topic.
Topic code: [for example EXOP]
Canonical topic name: [for example Example Co Operations]
Tool and tool class: [for example Slack, chat]
Purpose of the surface: [what it is for]

Give me:
1. The exact name of the surface, following the TFW naming rule for this
   tool class and the product's naming limits.
2. The description or topic text to set on the surface, with the purpose,
   the audience, and where the source of truth is.
3. The permission settings that match one access boundary for the topic.
4. Whether the topic already should have a surface of this class. If it
   should, tell me to repair or reuse that surface instead of creating a
   second one.

Decide where something belongs

Read https://tfw.how/llms-full.txt. It describes Topic Focused Workflow (TFW).
Use its terms exactly as defined in its glossary.

Our topics and topic codes: [list each topic code and canonical name]

Decide where this item belongs:
[Describe the item: a decision, a file, a task, a procedure, a message, and so on.]

Answer with:
1. The most specific valid topic, or a registry gap if none matches.
2. The kind of information (for example task, decision, file, procedure).
3. The source of truth: the tool class and the surface of that topic.
4. Which other surfaces should link to it, and what short link or note to
   post in the conversation where the item came from.

Skills

Agent Skills package instructions so that an AI agent can apply a method without a long prompt. Agent Skills for TFW, for Claude, ChatGPT, and other agents, are in preparation. They are not public yet. This page will link to them when they are published. Until then, use the prompts above with llms-full.txt.