The build skill

Case study · AI Tools & Strategy · Learning Design

The skill orchestrates AI agents cast as subject-matter experts to construct a course from the outcomes up, holding it to every online-quality standard on every pass. Where a quality audit checks a finished course against those standards, this skill enforces them while the course is being written, and runs the agents automatically.

The agents the skill orchestrates

The standards checklist

• Measurable objectivesQuality standardsWrites each outcome with an observable action verb at the appropriate level of the revised Bloom’s Taxonomy (Anderson & Krathwohl, 2001), the framework that orders learning from remember and understand, through apply, analyze, and evaluate, up to create. It flags unmeasurable verbs (understand, know, appreciate, be introduced to), noting the trap that “understand” is a valid level of the taxonomy but an invalid objective verb, and draws replacements from a curated verb bank rather than generating them ad hoc.
• Backward designStarts from the outcomes and the evidence that would prove them, then designs activities back from there, so the course is built in the order that keeps it coherent.
• Outcome → assessment alignmentBuilds the alignment matrix as it goes: every outcome tied to the graded evidence that substantiates it, nothing claimed that is never measured, nothing graded that carries no outcome.
• Seat hours and student workloadAccreditationEstimates each module, writes the total into the module intro, and holds the whole course to its credit-hour budget.

• Course orientation and getting startedQuality standardsA Welcome and Getting Started module: how to move through the course, where everything lives, what to do first, a structured introduce-yourself activity, and links to student support (advising, tutoring, accessibility, basic-needs).
• Syllabus structureA findable, printable syllabus built with every element the university requires: policies, modality, technical requirements, grading, and instructor contact. The course’s promise to the student.
• Instructor introduction and presenceA real introduction to the person teaching, so the course does not read as authored by no one. Presence is designed in from the first page.
• Clear expectations, for students and facultyParticipation and communication norms for students, including netiquette, alongside the instructor’s own commitments in writing, such as email response time and feedback turnaround, so both sides know what to expect.

• Universal Design for LearningMore than one way to demonstrate learning, content in more than one modality, and meaningful choice. UDL is a design framework, not accessibility; the skill keeps them separate.
• AccessibilityReal alt text on images, headings a screen reader can move through, links that say where they go, and tables assistive technology can read. The structural basics, since that is where most problems hide.
• Design and layoutQuality standardsConsistent navigation, readable contrast, clean heading structure, and clear instructions, never an undifferentiated block of text.
• Technology and toolsQuality standardsRequisite tech skills stated and scaffolded, no unused tools left in the menu, and a privacy-policy link for every third-party tool.
• Content and activitiesQuality standardsA variety of resources, higher-order thinking, authentic activities, OER and low-cost materials, and copyright and licensing stated throughout.

• Regular and Substantive InteractionFederal requirementInstructor presence, class community, and learner-to-learner collaboration built into every week, with stated expectations for every channel. A federal requirement for online courses, not an optional nicety.
• Communication cadenceA weekly rhythm of announcements, office hours, and check-ins, written into the course so the instructor inherits a plan rather than a blank calendar.
• Discussion boards and collaborationDiscussion prompts and collaborative activities designed for genuine engagement, with post-by and reply-by cadence, not a checkbox forum.
• FeedbackWhere and how feedback is returned, and on what timeline, so the loop between work and response is designed, not improvised.

• Authentic, AI-resistant by designEvery graded item faces one question: could an AI do this, correctly, in under a minute? If yes, it is rebuilt at the same cognitive level and engineered to collect process evidence (physical capture, critique given, revision from a named critique, articulated rationale, situated content), so the record of the making is the proof. No recall quizzes, no generic reflection prompts.
• Professional simulationsSummative assessment as a named professional scenario the learner defends on their own data or their own seed, so the same task cannot be reused or pasted from a model.
• The alternate-assessment menuDiscipline-aware options that replace a recall task with an apply-or-higher one at the same cognitive level, so rigor is kept while AI-substitutability drops.
• Rubric on every graded itemQuality standardsA grading and late-work policy, a rubric on every graded item, self-check opportunities, and a working gradebook that reconciles with the syllabus.
• The course-grounded AI tutorA public NotebookLM loaded with this course’s materials and its open resources, answering with citations. It supports the learner; it does not replace the instructor, and its answers are not graded interaction.

• Style guide appliedQuality standardsAt the end, one defined palette of solid colors (never gradients) and one typeface are applied across the whole course so it reads as designed, with contrast checked so nothing lands unreadable.
• Instructor prep kit and slide decksHanded to the faculty subject-matter expert with speaker-note scripts, a recording guide, and a weekly RSI rhythm.
• Course-evaluation survey and metricsAn end-of-course improvement survey, seat-time validation, and a place to measure learning, so the course can be improved on evidence rather than impression.
• The reference libraryEvery standard, verb bank, rubric, and framework the build drew on, kept with the course so the next revision starts from the same foundation.