Task decomposition
Separate complex work into bounded stages with clear inputs, outputs, dependencies, reviewers, and stopping points.
Unit overview
Unit 5 extends prompt fundamentals into context engineering and staged workflow design. You will determine what information belongs in context, which sources are permitted and authoritative, how a complex task can be decomposed, and when a prompt sequence is more appropriate than a single request. You will apply the author-approved SCOPE framework to connect sources, context, outputs, process stages, and evaluation or decision gates. Throughout the unit, AI output remains provisional and consequential decisions remain human responsibilities.
Context and source control
Provide necessary information, establish source hierarchies, and exclude unauthorized, irrelevant, or excessive material.
Prompt sequences and gates
Connect stages through review, verification, correction, approval, escalation, and final human decisions.
Central idea: More context and more prompts do not automatically
improve a workflow. Each stage should have a legitimate purpose, controlled information, an observable product,
and an accountable review gate.
Unit learning outcomes
By the end of Unit 5, you should be able to:
- Distinguish prompt engineering, context engineering, task decomposition, prompt chaining, iteration, and workflow design.
- Decompose a complex task into bounded stages with explicit inputs, outputs, dependencies, and human decisions.
- Identify the minimum context required for each stage and exclude unnecessary or unauthorized information.
- Develop a source hierarchy that distinguishes authoritative, supporting, prohibited, and unverified sources.
- Design a sequence in which each prompt has a specific purpose and produces a reviewable intermediate product.
- Prevent unverified content from being treated as established input in later workflow stages.
- Compare alternative prompts, contexts, sources, and outputs using observable evaluation criteria.
- Define review, verification, approval, escalation, fallback, and stop gates.
- Apply the author-approved SCOPE framework to a bounded AI-supported workflow.
- Document prompt versions, context decisions, source use, outputs, corrections, and human decisions.
- Identify when workflow complexity or consequences make AI use unsuitable.
- Prepare advanced prompt and workflow evidence for Assignment 2 and later applied projects.
Unit 5 lessons
Complete the five lessons in order. The filenames are local to the dedicated Unit 5 folder.
5.1 Task Decomposition and Workflow Boundaries
Break a complex task into manageable stages and identify dependencies, human-only activities, consequential decisions, and stopping points.
5.2 Context Engineering and Information Minimization
Build stage-specific context, distinguish necessary from excessive information, control persistence, and preserve privacy and authority.
5.3 Source Hierarchies and Evidence Boundaries
Prioritize authoritative sources, separate evidence from generated content, manage conflicts, and prevent citation or source laundering.
5.4 Prompt Sequences, Iteration, and Review Gates
Design multi-stage prompts with intermediate products, verification, correction, approval, escalation, fallback, and stop conditions.
5.5 SCOPE Workflow Design and Evaluation
Apply the author-approved SCOPE framework, evaluate the complete workflow, and preserve advanced portfolio evidence.
Folder convention: Keep
unit-05.html and t-01.html
through t-05.html together in a dedicated Unit 5 Moodle folder or content package.Embedded Unit 5 study guide
This sequence supports study from late Week 3 through Week 4 and advances the Prompt Engineering Portfolio.
- Decompose
Replace one complex request with bounded stages. - Control context
Define information and source boundaries for each stage. - Sequence and review
Add intermediate products and human gates. - Evaluate and document
Apply SCOPE and preserve the workflow record.
Suggested study sessions
Total: 7–9 hours- Session 1: Retrieval and decomposition, 60–75
minutes.
Retrieve PCAFE-V, read Chapter 5, and divide a supplied complex task into bounded stages. - Session 2: Lesson 1, 70–85 minutes.
Define stage purposes, inputs, outputs, dependencies, human activities, and stop points. - Session 3: Lesson 2, 70–85 minutes.
Build a context inventory for each stage and remove unnecessary or unauthorized information. - Session 4: Lesson 3, 70–85 minutes.
Create a source hierarchy and identify verification, conflict, currency, and authority questions. - Session 5: Lesson 4, 75–90 minutes.
Design a prompt sequence with review, correction, approval, escalation, fallback, and stopping. - Session 6: Lesson 5 and portfolio development, 90–120 minutes.
Apply SCOPE, test or inspect the workflow, compare alternatives, and update Assignments 1 and 2.
Evidence to retain
- Original complex task and decomposition map
- Stage-specific context inventory
- Source hierarchy and evidence boundaries
- Prompt sequence
- Intermediate outputs or supplied records
- Review and decision gates
- SCOPE workflow map
- Workflow evaluation and correction record
- Assignment 1 reflection notes
Show the advanced workflow sequence
- Confirm the legitimate purpose and suitability of AI.
- Decompose the work into bounded stages.
- Assign one clear purpose to each stage.
- Define the necessary context and permissible sources.
- Specify the intermediate output and evaluation criteria.
- Add independent verification before uncertain content is reused.
- Assign the human reviewer and decision-maker.
- Add escalation, fallback, and stop conditions.
- Evaluate the complete workflow, not only the final output.
- Document versions, evidence, corrections, decisions, and remaining uncertainty.
Assessment milestones in Unit 5
Assignment 2: Prompt Engineering Portfolio
Weight: 15%
Started: Unit 4, Week 3
Develop now: Add context engineering, source hierarchy, prompt sequencing, SCOPE workflow, and comparative evaluation.
Recommended submission: End of Module 2, Week 4
Official Moodle due date and time: ____________________
Assignment 1: Journal continues
Record how task decomposition, context minimization, source control, intermediate verification, and review gates changed your process and confidence.
Recommended final submission: End of Module 6, Week 11
Official Moodle due date and time: ____________________
Do not show only the final prompt: Retain the task decomposition,
context and source decisions, sequence, intermediate outputs, comparisons, corrections, and human decisions that
explain how the workflow developed.
Show the Unit 5 portfolio-record checklist
- Complex task and legitimate purpose
- Task decomposition and stage purposes
- Stage-specific context decisions
- Source hierarchy and prohibited sources
- Prompt sequence and intermediate products
- Evaluation and verification gates
- Human approval, escalation, fallback, and stopping
- SCOPE workflow representation
- Alternative workflow comparison
- Corrections, final human decision, and required disclosure
Required learning activities
| Activity | Purpose | Expected product | Estimated time |
|---|---|---|---|
| Task decomposition | Separate a complex task into bounded, reviewable stages. | Stage map | 35–50 minutes |
| Context inventory | Classify information as required, optional, unnecessary, sensitive, or prohibited. | Stage-specific inventory | 35–50 minutes |
| Source hierarchy | Prioritize authoritative evidence and identify supporting, conflicting, and excluded sources. | Source map | 35–50 minutes |
| Prompt-sequence design | Connect staged prompts without converting unverified output into assumed fact. | Prompt chain | 40–55 minutes |
| Review-gate design | Assign verification, correction, approval, escalation, fallback, and stop authority. | Gate and responsibility record | 35–50 minutes |
| Workflow comparison | Compare single-prompt, staged, human-led, and non-AI alternatives. | Comparative evaluation | 40–55 minutes |
| SCOPE application | Integrate sources, context, outputs, process, and evaluation. | Complete workflow map | 45–60 minutes |
Participation options
Option A: Approved AI system
Test a bounded prompt sequence using synthetic, public, supplied, or authorized information and no consequential action.
Option B: Supplied workflow records
Use supplied contexts, sources, prompts, intermediate outputs, corrections, and review records.
Option C: Non-AI workflow design
Design and evaluate the same staged process using conventional tools and human-produced intermediate work.
All options assess the same decomposition, context, source, sequencing, evaluation, and responsibility outcomes.
Responsible-use boundaries and support
Boundaries for Unit 5
- Do not enter confidential, personal, proprietary, privileged, security-sensitive, or culturally governed information.
- Do not allow unverified generated content to become assumed context in later stages.
- Do not connect email, records, publishing, payment, or external-action tools.
- Do not treat retrieval as proof that a claim is supported.
- Do not make the workflow so complex that accountable human review becomes impractical.
Support and accessibility
- Use supplied workflow records if live AI is inaccessible, unavailable, or unwanted.
- Provide a text description for every visual workflow.
- Contact the course coordinator for assignment, source, or permitted-use questions.
- Use library, privacy, accessibility, or technical support as appropriate.