Prompt like an operator: Context, Task, Standards, Evidence
Give AI a clear task and check whether it followed your instructions.
In this lesson · F3 of 8
All lessons in this track →Progress for this track is saved in this browser only. No completion badge is issued. View your learning.
Start with the idea
Give AI a clear task and check whether it followed your instructions.
Say who the work is for, what to produce and what rules to follow. Give it the facts it may use. Good instructions help, but you still need to check the result.
CTSE: Context, Task, Standards and Evidence: the situation, the job, the rules and the source facts.
See the example and check your understanding
A made-up teaching example
A six-person team handles 40 requests a week, at 18 minutes each. That is 12 hours of current work. The AI trial has not happened. Those 12 hours are not proven AI savings.
Ask yourself
Can the draft say AI already saved 12 hours?
Show the explanation
No. It can propose a trial and explain what to measure. It cannot turn current workload into achieved savings.
Start with these checks
- Separate plans from measured results.
- Recalculate any numbers you use.
- Check the next draft after changing the prompt.
Use the different practice case below and check every requirement in its full instructions.
The video explains the main idea. Use the sources, instructions and checks below to complete the practice.
Captions appear here during playback.
Prompt like an operator: Context, Task, Standards, Evidence
Give AI a clear task and check whether it followed your instructions.
Video playback is unavailable. Read the descriptive transcript below, go directly to practice, or open the video file.
Source-based teaching example · Fictional data · Reviewed 2026-10-03. Microsoft Mark · locally synthesized. On a small screen, use fullscreen for diagram details or read the transcript.
Read the descriptive transcript
0:00 · What you will learn
Give AI a clear task and check whether it followed your instructions.
What appears on screen
Give AI a clear task and check whether it followed your instructions.
0:06 · The idea
Say who the work is for, what to produce and what rules to follow. Give it the facts it may use. Good instructions help, but you still need to check the result.
What appears on screen
Say who the work is for, what to produce and what rules to follow. Give it the facts it may use. Good instructions help, but you still need to check the result.
0:20 · One useful term
CTSE. Context, Task, Standards and Evidence: the situation, the job, the rules and the source facts.
What appears on screen
Context, Task, Standards and Evidence: the situation, the job, the rules and the source facts.
0:32 · A worked example · 1
A six-person team handles 40 requests a week, at 18 minutes each. That is 12 hours of current work. The AI trial has not happened.
What appears on screen
A six-person team handles 40 requests a week, at 18 minutes each. That is 12 hours of current work. The AI trial has not happened.
0:45 · A worked example · 2
Those 12 hours are not proven AI savings.
What appears on screen
Those 12 hours are not proven AI savings.
0:51 · Your turn to think
Can the draft say AI already saved 12 hours? Pause and decide before continuing.
What appears on screen
Can the draft say AI already saved 12 hours? Pause the video if you need more time.
1:08 · Why this answer works
No. It can propose a trial and explain what to measure. It cannot turn current workload into achieved savings.
What appears on screen
No. It can propose a trial and explain what to measure. It cannot turn current workload into achieved savings.
1:18 · Check 1
Separate plans from measured results.
What appears on screen
Separate plans from measured results.
1:24 · Check 2
Recalculate any numbers you use.
What appears on screen
Recalculate any numbers you use.
1:29 · Check 3
Check the next draft after changing the prompt.
What appears on screen
Check the next draft after changing the prompt.
1:35 · Now try a different case
Open the lesson and try its different practice case. Use every check in the full task. Watching alone does not mark the lesson complete.
What appears on screen
Open the lesson Try the new case Check and revise The full task and its requirements are in the linked lesson.
Watching explains the example. Complete the lesson practice to check what you can do. Explore all visual lessons →
Before you start
Prerequisite: F2 authorization and Foundations source audit. Offline/no-cost route: paper, a notes editor and the synthetic packet below. Optional tools must already be authorized. No accounts, paid calls, workplace records or external actions are required.
What you will be able to do
- Write a CTSE prompt.
- Specify format, constraints, examples, and abstention.
- Match evidence requirements to consequence.
Part 1 CTSE separates instructions from proof
Context names the audience, decision and relevant supplied facts. Task requests one bounded output. Standards declare format, prohibited claims, uncertainty behavior and acceptance. Evidence specifies source IDs, passages, calculations or tests that will support material claims. This extends the older CTS frame; adding Evidence makes checks visible. A prompt can request abstention, but cannot enforce filesystem or network permissions. Version prompts and change one important variable at a time so the comparison has an interpretable cause. Keep task facts separate from style examples: copying example achievements into a new draft is a factual error.
Part 2 Worked synthetic example
Synthetic packet S1: six-person team, 40 weekly requests, 18 minutes per request, no AI trial yet. Weak request: make this better. CTSE: audience is operations manager; draft three-sentence trial proposal; preserve facts and distinguish proposals from results; map claims to S1 and recompute baseline effort. 40 × 18 = 720 minutes, or 12 hours weekly of request processing. This is baseline workload, not achievable savings.
Part 3 Guided practice and feedback
Compare “AI saves 12 hours weekly” with “Current request processing totals about 12 hours weekly; trial savings are unknown.” The latter is supported. Revise only the Evidence clause to demand baseline/benefit distinction, then inspect the next draft. Better instructions reduce ambiguity but do not guarantee no errors.
Review and next step
Self-review demonstrates practice, not independent certification or operational approval. F4 develops the claim audit used here.
Substantive review: 2026-10-02. See the configuration guide. OWASP excessive-agency guidance discusses the risks of unnecessary capabilities, permissions and autonomy.