[curriculum-eval] curriculum_assessment.py: checkpoint_quality — checkpoint regex misses :white_check_mark: shortcode used corpus-wide

[github-actions]

File: .github/skills/curriculum-quantitative-assessment/curriculum_assessment.py (tooling bug affecting all 93 workshop files, root cause of all 5 rubric findings)
Overall Score: N/A — this is a scorer defect, not a content defect (corpus mean: 6.14 / 10.0)

Flagged Dimensions:

Dimension	Score	Benchmark	Delta
checkpoint_quality	0.0 (in 88/93 files)	10 if checkpoint present with ≥4 items	-10.0

Root Cause (≤ 2 sentences):
CHECKPOINT_RE = re.compile(r"##\s+✅\s*Checkpoint", re.IGNORECASE) in curriculum_assessment.py only matches a literal ✅ emoji character, but every workshop step (per repository convention, workshop/00-welcome.md and 87 other files) uses the GitHub Markdown shortcode :white_check_mark: in headings like ## :white_check_mark: Checkpoint. As a result, has_checkpoint is False and checkpoint_quality scores 0.0 for 88 of 93 files — including all 5 files flagged in findings (side-quest-01-02-environment-reference.md, side-quest-17-07-repo-poisoning.md, 04-github-actions-intro.md, side-quest-11-06-anthropic-key.md, 05-agentic-workflows-intro.md) — even though each of these files verifiably has a well-formed ## :white_check_mark: Checkpoint section with a specific, multi-item checklist directly beneath it.

Evidence (quoted from the file):

CHECKPOINT_RE = re.compile(r"##\s+✅\s*Checkpoint", re.IGNORECASE)

Compare against the actual heading used throughout the corpus (e.g. workshop/05-agentic-workflows-intro.md):

## :white_check_mark: Checkpoint

Learning Science Rationale:
This is a measurement-validity problem rather than a curriculum-design problem: the rubric's construct (verifiable checkpoints that support self-assessment and formative feedback, per mastery-learning theory) is present in the actual content, but the instrument fails to detect it due to a shortcode-vs-emoji mismatch. Reporting checkpoint_quality: 0 for pages that do have well-formed checkpoints produces false-negative findings, wastes reviewer/agent effort chasing a non-existent gap, and — because checkpoint_quality carries a 2.0 weight — depresses the overall_score of 88/93 files by roughly 1.5–2.0 points each, systematically skewing the corpus mean, the Part 1/Part 2 summaries, and every trend calculation in score-history.json.

Improvement Prompt (for an agent):

Open .github/skills/curriculum-quantitative-assessment/curriculum_assessment.py and fix
CHECKPOINT_RE so it matches the repository's actual checkpoint heading convention. The
current pattern is:

    CHECKPOINT_RE = re.compile(r"##\s+✅\s*Checkpoint", re.IGNORECASE)

Update it to also match the GitHub Markdown shortcode form used throughout workshop/*.md,
for example:

    CHECKPOINT_RE = re.compile(r"##\s+(?:✅|:white_check_mark:)\s*Checkpoint", re.IGNORECASE)

Re-run the rubric scorer against the workshop/ corpus and confirm has_checkpoint becomes
True for all 88 files using the ":white_check_mark:" heading style (verify with:
grep -rl "^## :white_check_mark: Checkpoint" workshop/*.md | wc -l  → should be 88).
Regenerate corpus-metrics.json, rubric-results.json, and score-history.json (or the
equivalent generation step) so downstream consumers see corrected checkpoint_quality,
overall_score, and corpus mean/stdev values. Do not modify any workshop/*.md content —
the checkpoints themselves are already correct; only the detection regex is wrong.

Expected Score After Fix: Corpus mean rises from 6.14 / 10.0 toward an estimated 7.5–8.0 / 10.0 once checkpoint_quality correctly scores the 88 affected files; each of the 5 originally flagged findings should move from below-threshold (4.93–5.43) to above the 5.52 threshold once checkpoint_quality is restored (e.g. side-quest-01-02-environment-reference.md from 4.93 to an estimated ~7.4).

Generated by 🔬 Curriculum Quality Evaluator · copilot · auto · 96.3 AIC · ⌖ 12.2 AIC · ⊞ 9.5K · ◷

• expires on Sep 18, 2026, 1:02 PM UTC