[workshop-sim] Improve: Reduce task-stacking cognitive load in 07-your-first-workflow.md

[github-actions]

Problem statement

07-your-first-workflow.md is the third-highest-dropout step in the curriculum (20.0% conditional dropout among 21,923 at-risk simulated runs, 95% MC interval 19.5%–20.6%), driven mainly by copilot-access-missing (3,792 failures) and workflow-authoring-friction (1,207 failures). Content inspection shows the page has the highest cognitive_load score in the corpus (10.0/10) because it stacks a Copilot-access verification task, a full prompt-engineering explanation (the /agentic-workflows skill prompt and its NOTE callout about agent-harness differences), YAML frontmatter review, a compile step, and a commit/push step all on one page — while the underlying compile/publish state itself is well-supported (confirmed via content evaluation: both Step 7's own commit/push section and the required Step 7d billing side quest walk learners through gh aw compile → git commit → git push at least twice before Step 8).

Proposed change

Move the "Use the AI agent..." explanatory NOTE callout and its harness-difference rationale into a linked callout/side-quest reference (keep only a one-line pointer inline), so the main page flow is: verify access → run the authoring prompt → review output → compile → commit/push → checkpoint. This reduces the amount of prose a learner must parse mid-task without removing the prompt template, the generated-file example, or any checklist item.

Failure mode classification

Access barrier for the copilot-access-missing failures (the page already gates hard on this with a clear "do not continue" callout and a link to Step 7d for troubleshooting) combined with a learning barrier contribution from workflow-authoring-friction (task/concept stacking). This repair targets the learning-barrier portion; the access-barrier portion is already well-handled by the existing hard gate and troubleshooting link and should not be weakened.

Quantitative guardrail

Current scores for 07-your-first-workflow.md:

• overall_score: 6.25/10
• Weakest rubric dimensions: checkpoint_quality 0.0/10 (rubric detection artifact — page uses :white_check_mark: shortcode) and readability 8.3/10 with cognitive_load at the corpus maximum (10.0/10)
• Learning KPI index: (2.0×6.3 + 2.0×0.0 + 1.5×5.0)/5.5 = 3.65/10 (already the second-highest KPI index among the three selected high-risk steps, reflecting genuinely strong hands-on task design)

Trimming the inline harness-difference explanation into a linked callout reduces cognitive_load without touching the hands-on authoring task, the generated-file example, the compile step, or the checklist — so active_learning and scaffolding should stay flat or improve, and overall_score/KPI index should not regress.

Acceptance criteria

• The shared curriculum quantitative assessment overall_score for 07-your-first-workflow.md stays flat or improves after the change (baseline: 6.25/10)
• The step's learning KPI index (2.0 × active_learning + 2.0 × checkpoint_quality + 1.5 × scaffolding) / 5.5 stays flat or improves (baseline: 3.65/10)
• The prompt template, generated-file example, compile step, commit/push step, and full end-of-page checklist are preserved unchanged — only the harness-difference explanation is relocated, not removed
• The existing hard gate on Copilot access ("do not continue" callout + link to Step 7d) remains fully intact and undiminished

Suggested owner

copilot coding agent
Related to #3152

Generated by 🔬 Workshop Student Simulator · copilot · auto · 271 AIC · ⌖ 27.2 AIC · ⊞ 15.2K · ◷

• expires on Sep 11, 2026, 3:18 AM UTC