D685 Practical Applications of Prompt, catalog number ICSC 2212, is a two CU course in the WGU School of Technology introducing generative artificial intelligence and the skills involved in writing effective prompts. One catalog quirk is worth flagging: the title is transcribed here exactly as printed in the institutional catalog, where it appears truncated in all six of its instances, so you may see it completed differently in other materials.
Specification, not incantation
The unhelpful way to think about prompting is as a set of phrases that unlock better output. The useful way, and the one ICSC 2212 assesses, is as specification writing. A prompt is an instruction to a system whose interpretation you cannot inspect, which means ambiguity in the instruction becomes variation in the result, and the discipline required is the same one that makes a good requirements document.
That reframing changes what a strong submission looks like. Instead of a collection of prompts that produced pleasing output, it contains prompts with stated objectives, defined output formats, explicit constraints, and evidence of what happened when the prompt was varied. The variation is the evidence. A prompt reported once, with one output, demonstrates nothing that could not have been luck.
The second scored theme is verification. Generative systems produce fluent text regardless of whether the content is correct, and a submission that accepts output at face value has missed the point of a technical course on the subject. What belongs in the work is a checking procedure: which claims were verified, against what, and what was found to be wrong.
Third, at two competency units this is a small course, and the commonest way to lose time in it is to treat it as trivial. The aspects still require evidence, structure and citation like any other.
Turning scored aspects into a section plan
Scoring detail lives in your Course of Study rather than the catalog. Count the aspects and use them as headings. Each is judged independently on a three point scale and each needs a 2, so impressive output attached to a thin evaluation section is a returned task.
Where an aspect is satisfied by a prompt and its output, include both in full rather than summarising, and label them so the narrative can point at them precisely.
The word budget, worked. A two CU course usually carries a shorter expectation, so assume five scored aspects and around 1,200 words of narrative alongside the prompt and output appendices. Reserve 100 words for an opening naming the task and 90 for a close. That leaves near 1,010, about 200 per aspect. Then take 40 words from each of two descriptive aspects and give 80 to the aspect covering iteration and evaluation. Showing how a prompt changed and why is the scarce material; showing a prompt is not.
A structure that fits a prompting deliverable
Where your directions specify a structure, follow it exactly. Where they leave it open, this ordering matches how the aspects tend to be scored.
| Section | What belongs in it | How it gets read |
|---|---|---|
| Objective | What the output must contain and what it will be used for | Without this, no prompt can be judged effective or otherwise |
| Success criteria | How you would recognise a good output before seeing one | Criteria set afterwards always seem to be met |
| Initial prompt | The full text, with the reasoning behind its structure | Structure includes role, context, task, format and constraints |
| Output and evaluation | The result, assessed against the criteria rather than by impression | Impression based evaluation is the standard weakness |
| Iteration | Each revision, what changed, and what effect it had | One change at a time, or the effect cannot be attributed |
| Verification | Which claims were checked, against what source, and what was wrong | The section that makes this a technical submission |
| Limits and ethics | What this approach cannot do, and what disclosure is required | Brief but expected |
Change one element per iteration. Rewriting a prompt entirely and reporting that the output improved tells you nothing about which change caused it, which is the same experimental discipline any technical course expects.
Evidence craft for prompting work
Because outputs vary between runs, evidence discipline matters more here than the subject's reputation suggests.
- Record the model, its version and the date, since behaviour changes and an undated result is uninterpretable.
- Run the same prompt more than once and report the variation, because a single lucky output is not evidence.
- Include prompts verbatim, including whitespace and formatting, since both affect results.
- Verify factual claims against a real source and report the verification, not just the conclusion.
- Cite the techniques you apply to published sources rather than to social media guidance.
- Use APA in the written portion and disclose your use exactly as the task directions require.
Report a prompt that failed and could not be rescued. Knowing the boundary of a technique is more useful than another success, and it satisfies the limitations aspect with something specific.
What separates Competent from a submission sent back
Aspects score independently, and returns here concentrate on evaluation and verification.
- Success criteria were written before any output was generated.
- Each iteration changes one thing and reports its effect.
- Outputs are assessed against criteria rather than described as good.
- Factual content is verified against an external source and the check is reported.
- Model, version and date appear with every result.
Performance assessment work can be revised and resubmitted with no grade penalty, so a return costs calendar. Terms run six months at a flat rate, and a two CU course that takes three weeks because of a missing evaluation section is a poor trade.
Five mistakes that cost time in D685
- Treating prompting as phrasing tricks. The assessed skill is specification, and specification has structure.
- Evaluating by impression. Output described as much better with no criteria satisfies no aspect.
- Rewriting everything between iterations. Attribution becomes impossible and the iteration section loses its value.
- Accepting factual claims. Fluent and wrong is the characteristic failure mode, and checking it is the technical content of the course.
- Underestimating a two CU course. Fewer competency units means less content, not a lower evidentiary standard.
- Putting sensitive material into a prompt. Anything pasted into an external service has left your control, and coursework is a good place to build the habit of stripping identifying detail before it becomes a workplace problem.
The anatomy of a prompt that can be assessed
A prompt written for coursework should be constructed from components you can name, because named components can be varied deliberately and discussed in the report.
Role establishes the perspective the response should be written from, which mostly affects vocabulary and assumed background. Context supplies the situation and any material the system must work from, and it is where most improvement actually comes from, since a well specified context usually beats a clever instruction. Task states what to produce, using a single unambiguous verb rather than a description of a general aim.
Format specifies the shape of the output: length, structure, headings, whether prose or a list, and any field the result must contain. This is the component students most often omit and the one that most reliably reduces variation between runs. Constraints state what must not happen, which is different from what should happen and often more effective, since prohibitions are easier to check than aspirations. Examples, where the task allows them, demonstrate the target rather than describing it and typically do more than any amount of additional instruction.
Writing prompts this way gives the report its structure for free. Each iteration can name the component it changed, the reason for changing it, and the measured effect against the criteria, which is exactly the evidence the aspects are looking for and is impossible to produce from a prompt written as one undifferentiated paragraph.
How support works on this course
Send the rubric from your Course of Study and the task directions. What comes back is aspect mapped: criteria written before generation, prompts built from named components, an iteration log that changes one thing at a time, and a verification procedure. Plus a walkthrough so the method is yours to apply in the courses that follow.
At two competency units D685 is one of the quickest courses to close, provided the evaluation discipline is in place from the start. Terms are six months at a flat rate, so small courses cleared promptly are what create room for the heavy ones.
Questions students ask about D685
Is D685 the same course as ICSC 2212?
Which model should I use for the coursework?
Can you complete the prompting task for me?
Where D685 sits in WGU's programs
The July 2026 catalog places this code in 5 current WGU programs. Open a program page for the complete standard path and term positions. The live Degree Plan remains authoritative after transfer credit, substitutions, and mentor planning.
The assessments, one by one
The public catalog does not publish this course's PA/OA identity or task count. WGU Tutors publishes at most one PA manual per course and only from a WGU-controlled public rubric. Until that source exists, PA help begins from the student's real Course of Study and OA support remains preparation only.