---
name: context-page-one
description: Compress scattered context for recurring AI-assisted work into one portable page a model can read cold, with facts kept separate from unverified claims. Use when the same task keeps needing repeated corrections or re-explaining.
license: All-Rights-Reserved
metadata:
  author: Zain Haseeb
  version: 1.0.1
  changed: "2026-09-08"
  category: ai-workflows
  tags: [context, portability, sources, reusable-work]
---

# Context Page One

Build a single page of context that works when pasted into a brand-new AI session with no memory of anything before it.

## Critical behavior

1. The page must be self-sufficient. Assume the model reading it has no other history.
2. Separate standing facts from a worked example from explicit exclusions, in that order.
3. Flag any fact not backed by evidence as unverified rather than stating it as settled.
4. Keep the whole page short enough to fit on one screen without scrolling past it repeatedly.
5. Do not add generic best-practice filler that is not specific to the actual task.
6. State a maintenance trigger: what would make this page stale.

## When to use

Use this when a recurring AI-assisted task keeps producing generic or inconsistent output because the real context, standards, and boundaries live scattered across someone's head, old messages, and a dozen half-remembered corrections. It is also useful before handing a task to a new team member or a new agent that will run it going forward.

Do not use it for a one-off task where building a durable reference costs more than it saves.

## Build the page

Before collecting or exporting context, confirm permission for both the source material and intended destination. Do not copy private messages, transcripts, customer data, personnel details, or contract terms into another session without authorization. Minimize and anonymize examples; omit secrets and identifying details. Record source and freshness using non-sensitive references rather than reproducing protected content.

### 1. State who and why

Write one or two sentences naming who the output serves and why it matters. This orients every later judgment call the model has to make.

### 2. List standing facts

List facts that stay true across most runs of this task: audience, constraints, terminology, standards. Mark any fact you were not given clear evidence for as unverified instead of stating it as certain.

### 3. Give one worked example

Include one authorized, source-backed example of good past output. If none is available, mark the gap or provide an explicitly labeled synthetic example with the user's approval; never imply invented text came from a real case.

### 4. Name exclusions

State what is off limits, out of date, or should never appear in output built from this page: retired terminology, deprecated processes, confidential material, prior mistakes that keep recurring.

### 5. Add a maintenance note

State what change in the underlying work would make this page stale, and roughly how often it should be reviewed.

## Output format

Return the finished page exactly as it should be pasted into a new session, in four labeled blocks: who and why, standing facts, worked example, exclusions. Follow it with a short maintenance note, separate from the pasteable page itself.

## Example

A marketing lead keeps re-explaining brand voice and forbidden claims to a new AI session every week. A page one built from this skill fixes that: one paragraph on audience and purpose, five standing facts about tone and legal limits, one strong example post, and three named exclusions including a claim legal flagged last quarter. The lead now pastes one page instead of re-explaining from memory each time.

## Troubleshooting

- If a block would be empty because the information does not exist yet, say so directly rather than inventing plausible filler.
- If the page keeps growing past one screen, cut generic material first and keep only what is specific to this exact task.
- If output stays inconsistent after using the page, the problem is likely a missing worked example, not missing facts.

## What not to do

- Do not invent standing facts, audience details, or supposedly real examples. Keep any approved synthetic example explicitly labeled.
- Do not present an unverified fact as settled to make the page look more complete.
- Do not let the page become a general knowledge base; it is a working page for one recurring task.