ci: pin the grader to a stable branch name

[lassebenni]

What I built

Repoints grade-assignment.yml at lassebenni/github-actions-fork/.github/workflows/auto-grade.yml@stable/hyf-auto-grade and rewrites the stale comment above it. Where a docs/portal_confirm.md scaffold exists, its sample date also moves from May to August to match the chapter.

How to review

One workflow file. Confirm the uses: line and the comment block; there is no logic change.

Across the canonical repos the pin had drifted three ways: weeks 11 and 14 used @fix/allow-unsafe-pr-checkout-for-autograde, week 12 used upstream @main (which cannot check out fork PRs under pull_request_target), and the c55 cohort mirrors had already moved to @stable/hyf-auto-grade. This makes all of them agree.

The old branch name reads as disposable, but its upstream PR (HackYourFuture/github-actions#4) was closed without merge, so the pin is permanent. Deleting or force-pushing that branch would break grading for weeks 11, 12 and 14 simultaneously.

How to run

stable/hyf-auto-grade points at the same commit (346f1ae8) as the old ref, and the old branch is still in place, so this is safe to merge in any order. Open any assignment PR and confirm the Grade Assignment check still runs and produces a score.

What reviewers should see (expected results)

• uses: line ends in @stable/hyf-auto-grade
• Comment no longer claims the pin is temporary
• Grade Assignment behaves exactly as before

Self-check

• Target commit verified identical before pushing
• New ref confirmed reachable via the contents API
• Old branch left in place so nothing breaks mid-migration
• Same change already merged on the c55 week-14 mirror
• Merged by a human (not me)

🤖 Generated with Claude Code

[github-actions]

📝 HackYourFuture auto grade

Assignment Score: 25 / 100 ❌

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

Test Details

=== Week 12 Autograder — Orchestrated Pipeline ===
  ✓ PASS  found dags/taxi_pipeline.py
  ✓ PASS  found tests/test_dag_integrity.py
  ✓ PASS  found requirements.txt
  ✓ PASS  found RUNBOOK.md
  ✓ PASS  found ASSIGNMENT_REPORT.md
  ✓ PASS  found AI_ASSIST.md
  ✓ PASS  Level 1: required files (20/20 pts)
  ✗ FAIL  dags/taxi_pipeline.py: raise NotImplementedError still present — the DAG is not implemented
  ✓ PASS  dags/taxi_pipeline.py: defines a @dag with tasks
  ✓ PASS  Level 2: DAG implemented (5/15 pts)
  ✗ FAIL  dags/taxi_pipeline.py: only 1/3 expected tasks found (ingest, dbt_run, dbt_test)
  ✗ FAIL  dags/taxi_pipeline.py: no >> dependency chain found
  ✓ PASS  Level 3: sequential tasks (0/20 pts)
  ✗ FAIL  dags/taxi_pipeline.py: no uvx found — plain 'dbt' crashes on Python 3.14 (see Chapter 4)
  ✗ FAIL  dags/taxi_pipeline.py: no 'retries' found in default_args
  ✓ PASS  Level 4: uvx dbt + retries (0/20 pts)
  ✗ FAIL  dags/taxi_pipeline.py: no {{ ds }} / logical_date / get_current_context found — partition must come from the run date
  ✗ FAIL  dags/taxi_pipeline.py: catchup=False not found — required for safe normal operation
  ✓ PASS  Level 5: parameterized runs (0/15 pts)
  ✗ FAIL  RUNBOOK.md: still a template (374 chars, 7 TODO marker(s)) — fill in all four sections
  ✗ FAIL  AI_ASSIST.md: still a template (301 chars, 4 TODO marker(s))
  ✓ PASS  Level 6: documentation (0/10 pts)

Score: 25 / 100  (passing: 60)  pass=false

Reminder: the shared-Airflow deploy, the green run, and backfill
idempotency are Target-tier items a teacher reviews by hand — a high
static score here is necessary but not sufficient for Target.