Fix two Level 2 scoring holes and the PR template's deploy command

[lassebenni]

What I built

Two grader scoring holes and three PR-template errors, all reproduced against real trees before and after the fix.

Grader (.hyf/test.sh)

• Level 2 awarded 40 raw points into a 35-point cap, so any single baseline miss cost nothing. Rebalanced to 20 baseline + 15 extensions = 35 exactly, cap removed.
• The curated check passed on the string alone. Now requires a second container resource and the name.

PR template

• dummySecret= did not exist; the starter declares dbAdminPassword. Also added environment=, with a note that it only applies after Task 1.
• Described one container although Task 2 requires raw + curated.
• Self-check now references the teardown record.

Scaffold

• docs/portal_confirm.md prompts for the teardown line the assignment chapter now asks for.

How to review

• Read the rebalanced point comment at the top of Level 2 in .hyf/test.sh, then the curated block.
• Confirm the cap (if [[ "$l2" -gt 35 ]]) is gone and the raw sum is exactly 35.
• Skim the PR template diff for the parameter names.

How to run

git switch fix/grader-scoring-holes
bash .hyf/test.sh          # fresh starter: 40/100, Level 2 20/35

To reproduce the holes on main: delete the @secure() block from a solved tree and run the grader (scores 100/100 before this PR, 95 after); or append // TODO: add a container named 'curated' to a fresh modules/storage.bicep (awards the extension before, not after).

What reviewers should see (expected results)

Tree	Before	After
Fresh starter	40/100, L2 20/35	40/100, L2 20/35 (unchanged)
Solved	100/100	100/100
Solved, @secure() deleted	100/100	95/100
Solved, @secure() + module deleted	95/100	90/100
Starter + 'curated' in a comment	L2 30/35	L2 20/35

The fresh-starter number is deliberately unchanged, because the assignment chapter publishes it.

Known limitations / out of scope

• Grading still runs from lassebenni/github-actions-fork@fix/allow-unsafe-pr-checkout-for-autograde, a personal fork at a non-default ref, whose upstream PR was closed without merge. Not touched here, but it is a single point of failure for cohort grading.
• The grader stays static; it cannot verify the live Azure deploy or the teardown, only that the teardown line was written.

Self-check

• bash .hyf/test.sh passes on a solved tree (100/100) and fails a fresh starter (40/100)
• All five scenarios above re-run against the patched grader
• bash -n .hyf/test.sh clean
• PR template deploy command verified against a real deploy into rg-hyf-students
• No secrets added

[github-actions]

📝 HackYourFuture auto grade

Assignment Score: 40 / 100 ❌

Status: ❌ Not passed
Minimum score to pass: 60
🧪 The auto grade is experimental and still being improved

Test Details

=== Week 14 Autograder ===
  ✓ PASS  found main.bicep
  ✓ PASS  found modules/storage.bicep
  ✓ PASS  found docs/deploy_succeeded.txt
  ✓ PASS  found docs/what_if.txt
  ✓ PASS  found docs/portal_confirm.md
  ✓ PASS  found WRITEUP.md
  ✓ PASS  found AI_ASSIST.md
  ✓ PASS  found README.md
  ✓ PASS  Level 1: required files (20/20 pts)
  ✓ PASS  main.bicep has an @secure() parameter
  ✓ PASS  main.bicep calls a module
  ✓ PASS  declares a storage account resource
  ✓ PASS  declares a nested blob container resource
  ✗ FAIL  main.bicep missing param environment (Task 1)
  ✗ FAIL  missing Environment tag wiring — pass tags from main into the module (Task 1)
  ✗ FAIL  need a second nested container named curated (keep raw; Task 2)
  ⚠ WARN  only 1 nested container resource(s) — Task 2 expects raw + curated (count >= 2)
  ✓ PASS  Level 2: Bicep baseline + extensions (20/35 pts)
  ✗ FAIL  docs/deploy_succeeded.txt still empty or missing Succeeded/provisioningState
  ✗ FAIL  docs/what_if.txt still looks like the scaffold placeholder
  ✗ FAIL  docs/portal_confirm.md is still the scaffold stub — write 2-3 sentences from the portal
  ✓ PASS  Level 3: deploy evidence (0/25 pts)
  ✗ FAIL  WRITEUP.md is still too short or still the scaffold stub
  ✗ FAIL  AI_ASSIST.md is still too short or still the scaffold stub
  ✓ PASS  Level 4: write-up + AI report (0/20 pts)

Score: 40 / 100  (passing: 60)  pass=false