Prompt Patterns

Context Loading

Supplies external information, state, constraints, and references the model lacks but must rely on to complete the task correctly.

Also known asGroundingRetrieval-Augmented Generation (RAG)In-Context Learning
Context Loading pattern diagram

Problem

The model cannot access private state, current facts, or internal policy unless provided at runtime, so it defaults to generic advice or hallucinated specifics.

Solution

Load required facts, constraints, and authoritative excerpts into context before asking for output (manually or via retrieval). This replaces guesswork with grounded evidence the model can condition on directly.

Example

One worked example. The book has 2–4 per technique.

Prompt
Context
Product: self-serve B2B support platform
Activation definition: "first shared inbox connected and first reply sent from the platform"

Current funnel (last 30 days)

* 12,400 trial signups
* 4,320 invited at least one teammate (35%)
* 2,140 connected a shared inbox (17%)
* 1,180 sent a first reply from the platform (10%)

Top drop-off reasons (from support tags)

* "Inbox verification taking too long" (31%)
* "Unsure which teammate roles to assign" (26%)
* "Confused by channel setup requirements" (22%)

Policy update (effective January 2026, excerpt)
"We do not import historical messages from a shared inbox until domain ownership is verified. Any workflow that previews or syncs message history before verification is non-compliant."

Constraints

* No major UI redesign this quarter
* Engineering capacity: 2 eng-weeks

Task
Write a proposal with four to six specific interventions that target the biggest drop-offs, and explain why each should move the funnel. Use the policy excerpt as a hard rule.
Response
1. Inbox verification expectation setting
   Add a lightweight verification-status message plus an email triggered when inbox connection starts, with expected timing and next steps.

2. Role setup checklist
   Add a short role-selection checklist before inbox setup so teams can assign owner, admin, and responder roles with less guesswork.

3. Channel setup guidance staged after verification
   Move inbox history import guidance to after inbox verification, link to channel-specific setup help, and stay compliant with the verification policy.

4. Recovery for stalled inbox connection
   Send a single reminder after 24 hours if inbox verification is still pending, with clear status and a support escalation path.

Techniques

Concrete ways to implement Context Loading. Each technique fits a different situation.

  • 01

    Context Stuffing

    Place the relevant facts, state, and authoritative excerpts directly in the prompt so the model reasons from supplied ground truth.

  • 02

    Context Retrieval

    Write a retrieval policy instructing the model when and how to fetch information via tools before answering.

Prompt Patterns book cover

Full treatment in the book

Context Loading — the complete chapter

  • Mechanism — why this pattern works
  • 2–4 worked examples per technique
  • Placement, sequencing, and debugging rules
  • Composition with related patterns