lab-retro — Claude Code skill: retrospective and setup
lab-retro guides Claude Code Lab participants through a progress review, a reusable prompt, a monthly plan, and feedback. Run the full retrospective or jump to a section.
MIT 182 installs ★ 414 updated 2026-10-23
npx skills add glebis/claude-skills --skill lab-retro -a claude-codeWhat lab-retro does
The skill asks structured questions with AskUserQuestion and saves four outputs in lab-retro-output/: a before-and-after skills review, a prompt formatted as a skill, a monthly plan, and feedback. It runs the sections in sequence, but you can start at a particular one with /lab-retro 2.
The plan lays out weekly goals, prompts, success criteria, and conditions for taking a pause. A high burnout-risk rating adds a mandatory rest day. At the end, the skill sends feedback to the organizer through a public proxy and reports the returned row ID; if the request fails, the files stay local. The submission step runs curl and jq and needs the participant’s name as an argument. Without those tools, that step will fail.
What it helps with
- compare skills before and after the lab
- turn a useful prompt into a reusable skill draft
- make a four-week learning plan
- save feedback for the organizer and submit it through the proxy
Example prompts for the agent
- Run
/lab-retroand take me through the full retrospective. - Start with the second section:
/lab-retro 2. - Use
/lab-retroto help me plan the month after the lab.
How to install lab-retro
1Skills CLI (Vercel)
npx skills add glebis/claude-skills --skill lab-retro -a claude-code2Manually: put the files in place
~/.claude/skills/lab-retro/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/lab-retro/
1Skills CLI (Vercel)
npx skills add glebis/claude-skills --skill lab-retro -a codex2Manually: put the files in place
.agents/skills/lab-retro/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add glebis/claude-skills --skill lab-retro -a cursor2Manually: put the files in place
.cursor/skills/lab-retro/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install glebis/claude-skills lab-retro2Manually: put the files in place
.github/skills/lab-retro/SKILL.md1Skills CLI (Vercel)
npx skills add glebis/claude-skills --skill lab-retro -a gemini-cli2Platform CLI
gemini skills install https://github.com/glebis/claude-skills --path lab-retroManually: put the files in place
.devin/skills/lab-retro/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add glebis/claude-skills --skill lab-retro -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about lab-retro
How do I install lab-retro in Claude Code?
Run npx skills add glebis/claude-skills --skill lab-retro -a claude-code. The skill file is placed at ~/.claude/skills/lab-retro/SKILL.md.
What does lab-retro do?
It guides participants through four retrospective sections and saves a progress review, a reusable prompt, a monthly plan, and feedback in lab-retro-output/.
Can I run just one section of lab-retro?
Yes. Specify its number, such as /lab-retro 2, to jump to that section.
Source text
SKILL.md · MIT
---
name: lab-retro
description: Final retrospective and self-assessment for participants of Claude Code Lab. Runs four sequential interactive parts — progress audit, best prompt, monthly plan, and feedback — using AskUserQuestion. Triggers on "/lab-retro", "lab retrospective", "claude code lab final", or after completing the 6-week Claude Code Lab cohort.
---
# Claude Code Lab — Final Retrospective
This skill walks a Claude Code Lab graduate through four sequential exercises that consolidate their learning, capture their best work, plan next steps, and collect structured feedback for the organizer.
## How to run
Default flow: run all four parts in order. The user can also jump to a specific part with `/lab-retro 2` (or just say "part 3").
Between parts, briefly summarize what just happened and ask "ready for part N?" so the user controls the pace.
All artifacts are saved into a single folder `lab-retro-output/` in the current working directory:
- `01-progress.md`
- `02-best-prompt.md`
- `03-month-plan.md`
- `04-feedback.json` + `04-feedback-report.md`
Create the folder if missing.
---
## Part 1 — Progress audit
**Goal:** help the participant see concrete before/after.
Use `AskUserQuestion`:
1. **"Что вы умели ДО лаборатории?"** (multiSelect)
- Работал с ChatGPT/Claude в браузере
- Использовал CLI инструменты
- Писал код
- Автоматизировал задачи
- Работал с API
2. **"Что вы умеете ПОСЛЕ?"** (multiSelect)
- Работаю с Claude Code ежедневно
- Настроил MCP-серверы
- Создал свой Skill
- Автоматизировал реальную задачу
- Задеплоил что-то в веб
- Использую субагентов
- Пишу и читаю CLAUDE.md осознанно
3. **"Сколько часов в неделю экономит Claude Code?"** (singleSelect)
- <2
- 2–5
- 5–10
- 10+
Then output a markdown table comparing before/after with skill levels (0–5) and save to `lab-retro-output/01-progress.md`.
---
## Part 2 — Best prompt
**Goal:** turn one prompt the participant is proud of into a reusable Skill.
Ask the participant: *"Скопируйте или опишите ваш самый полезный промт из лабы."*
Then `AskUserQuestion`:
1. **"Для какой задачи был промт?"** (singleSelect)
- Автоматизация рутины
- Создание контента/документации
- Анализ данных/исследование
- Прототипирование/разработка
- Личный workflow / PKM
2. **"Что сделало его эффективным?"** (multiSelect)
- Хороший контекст в CLAUDE.md
- Чёткие критерии успеха
- Разбиение на шаги
- Использование Skills/MCP
- Примеры в промте
- Ограничения и анти-критерии
Reformat the prompt as a proper Skill (frontmatter + body), suggest an `description` line that would trigger it, and save to `lab-retro-output/02-best-prompt.md`. Suggest where to put it (`~/.claude/skills/<name>/SKILL.md`).
---
## Part 3 — Month plan
**Goal:** concrete 4-week plan so momentum doesn't die after the cohort.
`AskUserQuestion`:
1. **"Главная рабочая задача на месяц?"** (singleSelect)
- Проект для клиента/работодателя
- Свой продукт/стартап
- Автоматизация текущих процессов
- Обучение и развитие навыков
2. **"Сколько времени в неделю готовы уделять?"** (singleSelect)
- 15–30 минут
- 1–2 часа
- 3–5 часов
- Каждый день
3. **"Какой риск выгорания вы оцениваете для себя?"** (singleSelect)
- Низкий — у меня устойчивый ритм
- Средний — иногда залипаю
- Высокий — уже узнал себя в красных флагах
If risk = high, **insert a mandatory rest day** into the plan and a recommendation to read the AI hygiene slide again.
Generate `lab-retro-output/03-month-plan.md` with:
- 4 weekly goals
- 1–3 concrete prompts per week
- Success criterion per week
- Stop-conditions (when to pause)
---
## Part 4 — Feedback for the organizer
**Goal:** structured feedback that goes back to the lab organizer.
`AskUserQuestion`:
1. **"Оцените лабу в целом (NPS)"** (singleSelect: 0–10)
2. **"Самая ценная встреча?"** (singleSelect: M01 / M03 / M05 / M07 / M09 / M11)
3. **"Самая ценная тема за все 6 недель?"** (multiSelect)
- Основы Claude Code
- Промтинг и контекст
- Архитектура и субагенты
- MCP / Skills / Hooks
- Agent SDK и деплой
- Evals и качество
- AI-гигиена
4. **"Что улучшить?"** (free text)
5. **"Главное препятствие, с которым вы столкнулись?"** (free text)
6. **"Согласны ли поделиться отзывом публично?"** (singleSelect: да / да-анонимно / нет)
Save TWO files:
- `lab-retro-output/04-feedback.json` — structured for the organizer
- `lab-retro-output/04-feedback-report.md` — human-readable summary for the participant
**Then submit to the public proxy** (no secrets needed):
```bash
curl -sS -X POST https://lab-feedback-proxy.vercel.app/api/feedback \
-H "Content-Type: application/json" \
-d "$(jq -nc --arg name "<participant name>" --slurpfile notes lab-retro-output/04-feedback.json '{name:$name, notes:($notes[0]|tostring)}')"
```
The proxy forwards to Baserow table 746002 with a server-side token. Response is `{"ok":true,"row_id":<N>}`. Confirm row ID with the participant.
If the request fails, fall back to local files only and tell the participant: "submit failed — your feedback is saved locally in `lab-retro-output/04-feedback.json`, send it to the organizer manually."
---
## Closing
After all four parts, print:
1. Where the four files live
2. One sentence: *"Ваш фреймворк — LOOP. Ваш план — в файле 03. Ваш следующий шаг — открыть его в понедельник утром."*
3. Suggest committing the `lab-retro-output/` folder to a personal repo or vault.
## Constraints
- Always use `AskUserQuestion` for structured questions — don't ask in plain text
- Never skip a part silently; if the user opts out, write a one-line stub in the corresponding file
- Russian by default (audience is Russian-speaking); switch to English if the user replies in English
- Don't be sycophantic in the feedback report — surface honest patterns
© glebis · MIT · glebis/claude-skills