Purpose: Summative evidence for CO-03 (evaluate AI-generated HTML/CSS) and CO-04 (tool roles and prompting context). Judgement: S / NS only.

Time guide: 45–60 minutes.

Assessment tasks

Task 1 — Tools in the build loop (CO-04)

Complete the table: for each tool, give one primary role in a static-site build loop (e.g. plan, generate, refactor, explain). Use your own words; tool names are examples—substitute only if your provider allows.

Tool	One primary role in your workflow
Cursor (or IDE-integrated assistant)
Claude (or similar chat assistant)
Gemini (or second chat assistant)
ChatGPT (or similar)

Task 2 — Limitations and mitigations (CO-03 / CO-04)

List four important limitations of cloud AI assistants for coding (e.g. accuracy, context, privacy). For each, give one practical mitigation you would use. Format: bullet pairs or a two-column list.

Task 3 — Checklist application (CO-03)

Snippet for review (do not change the snippet—review it as given):

<body>
  <div class="header">My Cafe</div>
  <div>
    <div class="big-title">Welcome</div>
    <p>We are open daily.</p>
    <img src="cake.jpg" />
  </div>
  <div class="footer">Contact: call us</div>
</body>

Write 6–10 sentences that:

1. Name at least three issues or risks (structure, accessibility, or semantics).
2. For one issue, state what you would ask AI to change in a follow-up prompt (be specific).

Marking guide (Satisfactory / Not Satisfactory)

Criterion	Satisfactory (S)	Not Satisfactory (NS)
Task 1 — Roles	All four rows filled; roles are distinct enough and plausible; not copy-paste of product marketing.	Empty rows, or identical vague roles for every tool.
Task 2 — Limits/mitigations	Four limitation/mitigation pairs; mitigations are actionable (e.g. verify, smaller prompts), not “be careful” only.	Fewer than four pairs, or mitigations missing or generic.
Task 3 — Review	≥3 valid issues (e.g. no landmarks, heading as div, missing alt, weak footer semantics); follow-up prompt is specific.	Fewer than three issues, or issues invented that ignore the snippet, or no concrete prompt.

Overall module judgement: S only if all rows are S. Otherwise NS.

What good looks like (examples)

Task 1 (roles — one row):

Cursor: Generate and refactor markup directly in index.html while the folder is open.

Task 2 (limitation + mitigation):

Hallucination — model invents plausible but wrong tags → mitigation: run Module 03 checklist on every save before deploy.

Task 3 (extract):

The snippet uses generic divs instead of landmarks — “Welcome” is a styled div, not an h1, so screen reader users miss the page topic. The cake image has no alt. The footer says “call us” but is not a link or phone link. Follow-up prompt: “Rewrite this body using header, main, footer; make Welcome an h1; add descriptive alt to the cake image; make contact a tel: link.”

Exemplar responses (for assessors)

Task 3 issues may include: no header/nav/main/footer; “Welcome” not an h1; missing alt on image; “Contact” not a link; document structure. Accept reasonable alternatives if justified.

Reminder: Tools are descriptive; do not penalise learners for using different but equivalent assistants if the table is completed thoughtfully.