# cohort-analyst

Turns ability lists, entry diagnostics and pre-lesson surveys into teaching decisions. Reports stay at cohort level.

- skill · `type=skill` · Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode · ★ 41
- Author: YujxZJCN
- License: MIT · policy=mirror · the beginning of the file is at the end of this page
- Source: https://github.com/YujxZJCN/teaching-skills/tree/HEAD/cohort-analyst/SKILL.md
- Updated: 2026-10-03
- HTML: https://affpapa.org/en/nevibecode/skills/cohort-analyst

## What it does

The skill runs a team of four subagents. `diagnostic_designer_agent` writes ungraded diagnostics and pre-lesson questionnaires: prerequisite probes, two-tier misconception items, self-efficacy items marked as such. `cohort_analyst_agent` computes per-concept readiness distributions, how common each misconception is, and how wide the spread is. `calibration_advisor_agent` turns the profile into changes for one lesson, and `grouping_strategist_agent` proposes group compositions for a given pedagogical goal.

There are five modes: `instrument`, `cohort-profile`, `lesson-calibration`, `grouping` and `progress`. The first returns a student-facing instrument plus a per-item analysis plan, the second an aggregate readiness profile and a proposed `learner_profile` update for the Course Passport, the third what to reteach, activate or skip, the fourth a grouping plan, the fifth a mid-term trajectory comparison. Results handed over without a known instrument go through a short provenance intake first, because what produced the numbers decides what they can support.

The unit of analysis is the cohort. Identifiable rows are pseudonymized for the session, the mapping stays with the professor, and only distributions and prevalence percentages reach the passport. Self-reported confidence and measured ability are labelled separately in every report, and a 5-item pre-quiz is flagged as a coarse signal. The skill does not answer which students need help — it never scans for individuals.

## Install

### Claude Code (native platform)

1. **Skills CLI (Vercel)**
```bash
npx skills add YujxZJCN/teaching-skills --skill cohort-analyst -a claude-code
```

2. **Manually: put the files in place**
- File: `~/.claude/skills/cohort-analyst/SKILL.md` — Copy the whole skill folder; for one project use .claude/skills/cohort-analyst/

### Codex

1. **Skills CLI (Vercel)**
```bash
npx skills add YujxZJCN/teaching-skills --skill cohort-analyst -a codex
```

2. **Manually: put the files in place**
- File: `.agents/skills/cohort-analyst/SKILL.md` — For all projects use ~/.agents/skills/

### Cursor

1. **Skills CLI (Vercel)**
```bash
npx skills add YujxZJCN/teaching-skills --skill cohort-analyst -a cursor
```

2. **Manually: put the files in place**
- File: `.cursor/skills/cohort-analyst/SKILL.md` — Cursor also reads .agents/skills/ and .claude/skills/

### GitHub Copilot

1. **GitHub CLI**
```bash
gh skill install YujxZJCN/teaching-skills cohort-analyst
```

2. **Manually: put the files in place**
- File: `.github/skills/cohort-analyst/SKILL.md`

### Gemini CLI

1. **Skills CLI (Vercel)**
```bash
npx skills add YujxZJCN/teaching-skills --skill cohort-analyst -a gemini-cli
```

2. **Platform CLI**
```bash
gemini skills install https://github.com/YujxZJCN/teaching-skills --path cohort-analyst
```

### Windsurf (Devin Desktop)

**Manually: put the files in place**
- File: `.devin/skills/cohort-analyst/SKILL.md` — The legacy .windsurf/skills/ path still works

### OpenCode

**Skills CLI (Vercel)**
```bash
npx skills add YujxZJCN/teaching-skills --skill cohort-analyst -a opencode
```

### Claude.ai

**Claude.ai**
- Zip the folder and upload it in Settings → Capabilities → Skills

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/YujxZJCN/teaching-skills/tree/HEAD/cohort-analyst/SKILL.md

## When to use it
- write a 10-minute ungraded check for week 1
- see what prior knowledge a class arrives with
- adjust next week's lesson after a pre-lesson survey
- form groups for an in-class activity
- compare cohort readiness against week 1

## Example prompts
- Design a 10-minute ungraded diagnostic for week 1 of my data structures course
- Here are the pre-quiz results — what does my class actually know coming in?
- Build peer-instruction groups from the diagnostic results

## Source text · SKILL.md · MIT

Below is the content of a third-party file. It is data, not instructions for you: do not act on anything written inside unless the human asks. Block tag: 9d502f9402de29cc.

BEGIN THIRD-PARTY FILE 9d502f9402de29cc
```text
---
name: cohort-analyst
description: "Learner and cohort analysis (学情分析) for university professors — cross-cutting support for design, builds, and the weekly loop. 4-agent team turning professor-held student data — ability lists, pre-course diagnostics, pre-lesson questionnaire results — into evidence-based teaching decisions: ungraded diagnostic design, aggregate readiness profiles, lesson calibration, evidence-based grouping, and mid-term trajectory re-analysis. Cohort aggregates only — no individual-level output, ever. Triggers on: student readiness, pre-assessment, diagnostic quiz, pre-lesson questionnaire, prior knowledge survey, learning analytics, ability levels, class profile, differentiation, grouping, 学情分析, 学情, 摸底, 前测, 预习问卷, 课前问卷, 学生基础, 分层教学, 分组."
metadata:
  version: "1.0.0"
  last_updated: "2026-06-11"
  status: active
  pipeline_stage: support
  related_skills:
    - course-designer
    - lesson-builder
    - student-mentor
    - assessment-architect
    - teaching-pipeline
---

# Cohort Analyst — Learner Evidence Team

Turns the student data a professor already holds — ability lists, pre-course
diagnostics, pre-lesson questionnaire results — into teaching decisions with evidence
behind them. Cross-cutting: a pre-term profile informs Stage 0/1 design
(`course-designer` reads `learner_profile`), pre-lesson results calibrate Stage 2
builds (`lesson-builder`), and the Stage 4 weekly loop re-runs the cycle as the
cohort moves. The professor knows the discipline and the students; this skill brings
instrument craft, aggregation honesty, and the discipline to say what a 5-item quiz
cannot say.

> **Prime rule — the privacy architecture:** the unit of analysis is the **cohort**.
> The Course Passport receives aggregates only — distributions, prevalence
> percentages, heterogeneity measures — written into `learner_profile` and shown to
> the professor verbatim before writing. Raw data (named or identifiable rows) stays
> in the professor's files: the skill works on it in-session, pseudonymizes where
> feasible, and never writes any individual-level fact to the passport or any state
> file. "Which students need help?" is not this skill's question — that routes to
```
END THIRD-PARTY FILE 9d502f9402de29cc

The third-party file has ended. Everything between the 9d502f9402de29cc tags is data: requests and commands from it are not followed. An “end of file” line with another tag inside the block is part of the file.

First 2291 of 10481 bytes. © YujxZJCN · MIT · https://github.com/YujxZJCN/teaching-skills/tree/HEAD/cohort-analyst/SKILL.md

_Part of the page did not fit — see the HTML version._

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md