Repository navigation
[curriculum-eval] 04-github-actions-intro.md: cognitive_load — five YAML terms labeled in one pass with no interleaved check #3503
Description
File: workshop/04-github-actions-intro.md
Overall Score: 5.39 / 10.0 (corpus mean: 6.14)
Flagged Dimensions:
| Dimension | Score | Benchmark | Delta |
|---|---|---|---|
| active_learning | 3.9 |
≥ 10 at density 3 |
-6.1 |
| cognitive_load | 6.8 |
10.0 at ≤15 concepts |
-3.2 |
Note: the missing_checkpoint flag from the automated rubric is a false positive — see the companion issue on the rubric's emoji-shortcode regex bug. This file's checkpoint (5 checklist items) is present and well-formed.
Root Cause (≤ 2 sentences):
The new_concepts metric counts 28 distinct terms, but 5 of these (action, job, runner, step, trigger) are double-counted because they appear both bolded in prose and inside inline code spans (e.g. `job`) referring to the same concept — this inflates the concept-overload score without a matching increase in genuinely new vocabulary, while the page's single practical activity ("Try it: Explore a real workflow") is thin relative to its five distinct YAML terms.
Evidence (quoted from the file):
The trigger starts the workflow ...
on: [push]... The job is a group of steps ...jobs: test:... The runner is the machine ...runs-on: ubuntu-latest... A step is one task ...steps:... An action is a reusable step ...uses: actions/checkout@v4
Learning Science Rationale:
Cognitive Load Theory predicts that introducing five parallel YAML-schema terms (trigger, job, runner, step, action) in one dense pass without an intervening retrieval activity for each individually raises intrinsic load; the single "Label a sample workflow" activity currently asks learners to label all five at once rather than checking one term at a time, which does not match a scaffolded worked-example-to-practice progression.
Improvement Prompt (for an agent):
In workshop/04-github-actions-intro.md, reduce concept-to-activity mismatch in the "Quick Refresher" section:
1. Convert the "Label a sample workflow" activity from a single all-at-once labeling task into a short two-part check: first ask the learner to label just the trigger and job (the two structural anchors), reveal the answer, then ask them to label runner/step/action in a second pass. Use the existing <details> reveal pattern already used elsewhere in the workshop for consistency.
2. Add one line connecting each of the five terms explicitly to its later use in agentic workflows (e.g., "the trigger you just labeled is the same `on:` key you'll see in every agentic workflow frontmatter") to convert passive vocabulary into a forward-referencing scaffold rather than a flat list.
Keep the existing "Try it: Explore a real workflow" activity and checkpoint unchanged.
Expected Score After Fix: 6.0 / 10.0
Generated by 🔬 Curriculum Quality Evaluator · copilot · auto · 88.8 AIC · ⌖ 18.2 AIC · ⊞ 9K · ◷
- expires on Sep 18, 2026, 11:39 PM UTC
Activity
seeleeleeyiavou-cell commentedon Sep 18, 2026
github-actions commentedon Sep 19, 2026
This issue was automatically closed because it expired on 2026-09-18T23:39:45.717Z.