Steam Launch Check (countdown, refunds, AI disclosure, retro)
/steam-launch-check
A GameSkill original for the weeks around a Steam release: a countdown to every review deadline with business days and holidays, a refund-window check on playtest finish times, an item-by-item AI disclosure draft that rejects vague wording, a retro calculator for numbers exported from Steamworks, and review triage by playtime, version, topic and Steam Deck. Bilingual report, no dependencies, data stays local.
- 许可证
- MIT
- 出品
- GameSkill 出品
我们测过
在 Node 20 (no dependencies) + Chrome 156 headless for the report screenshot 上通过 33/33 项检查（2026-10-28）：能触发、有可运行示例、有截图和自检清单、单一职责、中英文文档齐全。触发检查是把提示词和 skill 描述做匹配，不是真实 agent 运行。
测试报告GameSkill 测过的版本
GameSkill 为 gameskill.dev 原创编写并测过。ZIP 里是完整的 skill：SKILL.md 及中文版、脚本、可运行示例、测试、自检清单和我们的测试报告。
GameSkill 原创，协议 MIT。ZIP 保留 LICENSE 和 CREDITS.md，里面列出了流程灵感来源、CC0 测试素材和第三方库及其协议。
自检清单
- 倒计时里的每个 Steam 截止日都还没过，或者已经说明最早可行发售日 / Every Steam deadline is ahead, or the earliest workable launch date was given
- 试玩通关中位数和 2 小时退款窗口比较过 / Playtest median finish time compared with the 2-hour refund window
- AI 披露逐项写了内容、数量、工具和人工处理，运行时生成的写了防护措施 / AI disclosure lists content, count, tool and human work per item, with guardrails for live content
- 复盘里的阈值标成参考值 / Retro thresholds presented as reference values
- 销售数据只在本地，没有转载评测原文 / Sales data kept local; no review text republished
- npm test 和 npm run test:cli 都通过 / npm test and npm run test:cli pass
推荐搭配
中文指南
这个 skill 做什么
GameSkill 出品（MIT），覆盖 Steam 发售前后几周：从发售日倒推每个审核截止日（上架费等待期、税务审核、提前 7 个工作日提交商店页审核、“即将推出”至少 2 周、构建审核 7 个工作日，自动跳过周末和你列出的节假日），并给出最早可行发售日；用试玩通关时长对比 2 小时退款窗口；逐项生成内容调查问卷的 AI 披露草稿，含糊的写法会被拦下；用 Steamworks 导出的数字做复盘（退款率、净收入占比、愿望单转化、游玩时长中位数）并给出可能原因；按游玩时长、版本、话题和 Steam Deck 给公开评测分类。输出中英双语报告，没有依赖，数据只在本地。
什么时候用
- 准备第一次在 Steam 发售，不知道什么时候该提交审核
- 游戏偏短，担心玩家 2 小时内通关就退款
- 要填内容调查问卷的 AI 部分
- 发售后想看懂销量、退款和评测
怎么装
下载本页的 ZIP，把 steam-launch-check/ 文件夹放进 .claude/skills/（Claude Code）或 .agents/skills/（Codex、Cursor），重启 agent。只需要 Node.js 20+，不用 npm install。
最小示例
node tools/launch-check.mjs report example 用我们自己的演示游戏《灯蛾》的虚构数据生成 out/report.zh.md、report.en.md 和 report.html：倒计时表、退款窗口告警（通关中位数 55 分钟）、通过检查的 AI 披露草稿、复盘表（退款率 25.2%，原因是退款窗口）和评测分类（免费 DLC 版本之后好评变多）。plan example/launch.json --today 2027-02-20 会报告已经来不及，并给出最早可行发售日 2027-03-17。
我们怎么测的
npm test 跑 9 个 Node 测试：工作日和节假日倒推、示例的每个节点日期、迟到检测和最早发售日、退款窗口三个等级、AI 披露的通过和三类错误、复盘的指标和原因、评测分组、CSV 解析、双语报告。npm run test:cli 做 10 项命令行检查：各命令的退出码和输出、中文倒计时、双语报告、示例数据都标了虚构，以及在无头 Chrome 里渲染报告截图。全部离线。
致谢
按版本、话题、Deck 和游玩时长给评测分类的思路来自 glitchless-gg/game-dev-skills（MIT），没有复制代码或文字。上架教训来自 @kunchevsky 的公开帖子和回复（只放链接，事实为转述，不使用他的游戏名、数据或截图）。Steam 规则来自 Steamworks 官方文档和 Steam 退款页面（2026-10-28 核对）。
搭配
game-trailer-from-code：预告片、capsule 和截图ccgs-launch-checklist：各平台上线清单performance-optimization：评测里“卡顿”话题多的时候
安装
告诉你的 AI
Install the skill from https://gameskill.dev/skills/steam-launch-check/install.md
GameSkill 测过的版本
mkdir -p .claude/skills
curl -fsSL https://gameskill.dev/dl/skills/steam-launch-check.zip -o /tmp/gameskill-steam-launch-check.zip
unzip -oq /tmp/gameskill-steam-launch-check.zip -d .claude/skills
rm /tmp/gameskill-steam-launch-check.zip包含内容
- tools/launch-check.mjs: plan, refund, disclosure, retro, reviews and report commands
- Countdown to the app fee wait, tax check, store and build review (business days) and Coming Soon, with the earliest launch that still works
- AI disclosure draft in English and Chinese with a vague-wording linter and live-content guardrails
- Retro and review triage with likely causes, labelled as reference values
- Fictional Lantern Moth example, 9 Node tests and 10 CLI checks
适合
- Indie teams planning a first Steam release
- Short games at risk of refunds inside the 2-hour window
- Filling in the Content Survey AI section without being sent back
不适合
- Trailers, capsules and screenshots: see game-trailer-from-code
- Console certification and other stores: see ccgs-launch-checklist
试试这条提示词
My game launches on Steam on 11 March 2027. When do I submit everything for review, and is my 50-minute game at risk of refunds?关于这个 Skill
A GameSkill original for the weeks around a Steam release. One dependency-free CLI turns a release date, playtest times, an AI-use list, numbers exported from Steamworks and the game's public reviews into a bilingual launch report, and exits with 1 when something needs action.
Agent steps
- Countdown: work back from the release date to the app fee wait, tax check, store review (7 business days before Coming Soon), Coming Soon (2 weeks) and build review (7 business days). If anything is late, give the earliest workable date instead of squeezing reviews.
- Refund window: compare the playtest median finish time with Steam's 2-hour window; flag short games before launch.
- AI disclosure: one item per use with exact content and count, tool, human work, and guardrails for live content; vague wording fails.
- Retro from day 14: refund rate on units sold before refunds, net vs gross, wishlist conversion, median playtime, with likely causes as reference values.
- Reviews: triage the user's own game by playtime, version, topic and Steam Deck; summarise, do not republish.
- Keep data local.
Common errors it fixes
- Counting calendar days when Steam's review windows are business days.
- Short games that players finish and refund inside 2 hours.
- A one-line "we used some AI" disclosure that gets the page sent back.
Credits
Review triage by version, topic, Deck and playtime is an idea from glitchless-gg/game-dev-skills (MIT); no code or text copied. Launch lessons from @kunchevsky's post and replies (link only, facts paraphrased). Steam rules from the Steamworks documentation and Steam Refunds page, checked 2026-10-28.
环境要求
- Node.js 20+
- Your own playtest times, Steamworks numbers and app ID (the example uses fictional data)
相关主题
常见问题
基于 glitchless-gg/game-dev-skills (review triage idea) （作者 glitchless-gg，MIT）的思路。代码和文字由 GameSkill 编写，ZIP 里的 CREDITS.md 列出所有来源。 · 基于 Steam launch post and replies (link only) （作者 @kunchevsky，no licence, facts paraphrased, nothing copied）的思路。代码和文字由 GameSkill 编写，ZIP 里的 CREDITS.md 列出所有来源。
搭配使用
作者标注的相关 Skills，或更适合相邻任务的 Skills。
Launch Checklist (Claude Code Game Studios)
Launch readiness across every department: code, content, store, marketing, community, infrastructure, legal, go/no-go sign-offs.
Game Trailer from Code (Steam trailer and store images)
A GameSkill original: turn your own running game into a 15-second Steam gameplay trailer and every store image with code. Frame-exact capture, a JSON cut list, zh/en captions and SRT, a soundtrack synced to game events, an H.264/AAC encode, and a checker for Steam's trailer rules, loudness, the 6-second microtrailer and the capsule sizes.
相似 skills
skills-gamedev (Skillary)
26 game development skills from the Skillary catalog: game design, level and narrative design, monetization, playtesting, Unity, Unreal, Blender, audio and engineering.
Playtest Report (Claude Code Game Studios)
Structured playtest report template, or turn existing playtest notes into structured feedback.
Universal Modder
Skills, tools and a fal MCP that let Claude Code mod PC games you own: recon, reverse engineering, generated assets, in-game testing.
Game Performance Optimization
Fix performance methodically: profile first, set a frame budget, find the CPU or GPU bottleneck, then pool, batch and cut allocations.