assessment-architect
Builds assessment instruments for university professors — pipeline Stage 3. 7-agent team covering test blueprints, exam/quiz/question-bank item writing, rubric design (analytic/holistic/single-point), TILT project briefs, AI-era integrity auditing, verified answer keys with worked solutions, and post-exam item analysis. Triggers on: exam, midterm, final, quiz, test questions, rubric, grading criteria, project brief, assignment design, question bank, item analysis, academic integrity, AI-proof, 出题, 试卷, 考试, 测验, 评分标准, 评分量表, 课程项目, 题库, 试题分析, 学术诚信.
YujxZJCN · Static checks on file
Installation
Read the source and dependencies before running a command. Copying does not install or execute anything.
npx skills add YujxZJCN/teaching-skills --skill assessment-architectAbout this skill
name: assessment-architect description: "Builds assessment instruments for university professors — pipeline Stage 3. 7-agent team covering test blueprints, exam/quiz/question-bank item writing, rubric design (analytic/holistic/single-point), TILT project briefs, AI-era integrity auditing, verified answer keys with worked solutions, and post-exam item analysis. Triggers on: exam, midterm, final, quiz, test questions, rubric, grading criteria, project brief, assignment design, question bank, item analysis, academic integrity, AI-proof, 出题, 试卷, 考试, 测验, 评分标准, 评分量表, 课程项目, 题库, 试题分析, 学术诚信." metadata: version: "1.0.0" last_updated: "2026-06-10" status: active pipeline_stage: 3 related_skills: - course-designer - lesson-builder - teaching-pipeline
Assessment Architect — Instrument Construction Team
Builds the actual instruments the assessment plan promised: exams, quizzes, question
banks, rubrics, project briefs — then audits them for AI-era integrity and analyzes how
they performed. The plan (types, weights, timing) is Stage 1 work and lives in the Course
Passport; this skill turns each assessment_plan entry into something students can sit,
submit, and be fairly graded on.
Prime rule: blueprint before items (Pedagogy Foundations §10). A test written question-by-question measures whatever was easy to ask; a test written from a confirmed content × Bloom blueprint measures the outcomes. No item is drafted before the professor confirms the blueprint — ever.
Quick Start
Write the midterm for CS 201 — it's A1 in the passport
给我的数据结构课出一份期末试卷
Build a rubric for the term paper
Design the semester project brief, AI-disclosure tier
我的考试结果在这个表里,帮我做试题分析
Audit my take-home final for AI vulnerability
Close my gradebook — show me what the A/B cutoff choices do
Design the group project with peer assessment
Make the 1.5×-time version of the midterm for a granted accommodation
Modes
| Mode | Trigger intent | Output |
|---|---|---|
exam | "Write the midterm/final/exam for…" | Blueprint → items → verified key → logistics (versions, accommodations, instructions) |
quiz | "Quiz on this week's material"; low-stakes retrieval | Short retrieval set with key, sized to minutes available (Pedagogy Foundations §5) |
question-bank | "Build a question bank / pool for…" | Tagged item bank with parallel variants for reuse and randomization |
rubric | "Rubric / grading criteria for…" | Analytic, holistic, or single-point rubric + TA calibration notes |
project-brief | "Design the project / assignment for…" | Student-facing TILT brief (Purpose/Task/Criteria) with milestones + instructor block |
integrity-check | "Is this AI-proof?"; pipeline Stage 3 audit | AI-resilience audit per shared/ai_era_integrity.md — standalone or pipeline, read-only |
item-analysis | "Analyze my exam results"; post-exam | Difficulty, discrimination, distractor analysis from a professor-provided results table |
answer-key | "Make/check the key for this exam" | Regenerated key for an existing instrument: worked solutions, grading notes, discrepancy flags |
grade-analysis | "Close my gradebook"; "what if the A line is at…" | Final-grade distribution + shape diagnostics, a what-if cutoff/curve comparator (counts only), fairness note — aggregates only; the professor sets cutoffs |
group-assessment | "Design the group project + peer assessment" | Graded group brief with genuine interdependence, individual-accountability mechanism, and a contribution-adjusting peer-assessment instrument |
accommodate | "Make the 1.5×-time / alt-format version of this exam" | Modified assessment material for an already-granted accommodation, equivalent rigor preserved + logistics note |
Mode dispatch rule: an instrument request that names no passport assessment runs standalone — intake the context, build, and offer passport write-back at exit (Passport Iron Rule 5). Detect intent in any language.
Does NOT trigger
| Scenario | Use instead |
|---|---|
| Deciding assessment types, weights, or timing (the structure) | course-designer |
| Practice activities and exercises that aren't graded | lesson-builder |
| Writing feedback comments on a specific student's work | student-mentor |
| Full design → materials → assessment run | teaching-pipeline |
| Deciding whether an accommodation is granted / who is eligible | disability/accessibility office (skill operationalizes an already-granted one) |
| Accommodating lecture, slide, or reading materials (not assessments) | lesson-builder / deck-studio |
| Analyzing a named borderline student's grade case | student-mentor (grade-analysis is cohort aggregates only) |
Agent Team (10)
| Agent | Role |
|---|---|
blueprint_agent | Builds the test blueprint: content × Bloom matrix from passport outcomes, point and time budgets; flags level mismatches |
item_writer_agent | Drafts items per blueprint cell using references/item_writing_rules.md; distractors from known misconceptions; bank variants |
rubric_designer_agent | Designs analytic/holistic/single-point rubrics with observable descriptors and TA calibration anchors |
project_designer_agent | Writes TILT-structured project and assignment briefs with milestone staging and scope honesty |
integrity_auditor_agent | Runs the shared/ai_era_integrity.md audit procedure; read-only; sets ai_resilience; never recommends detectors |
answer_key_agent | Independently works every item to produce the key + worked solutions; flags discrepancies, marks [VERIFY] where uncertain |
item_analyst_agent | Post-exam statistics: difficulty, discrimination, distractor performance; per-item action recommendations |
grade_analyst_agent | Closes the gradebook: weighted final-grade distribution, shape diagnostics, what-if cutoff/curve comparator, fairness note — aggregates only; never sets c |