launch-tweet
关于
This Claude Skill drafts concise launch tweets or short threads for solo product launches on X/Twitter. It's triggered when developers ship a real product with a live link and request launch announcements. The skill focuses on creating honest, substance-driven posts rather than engagement-bait, avoiding vague teases or inappropriate contexts like B2B launches.
快速安装
Claude Code
推荐npx skills add rockscy/solo-skills -a claude-code/plugin add https://github.com/rockscy/solo-skillsgit clone https://github.com/rockscy/solo-skills.git ~/.claude/skills/launch-tweet在 Claude Code 中复制并粘贴此命令以安装该技能
技能文档
Launch Tweet / 发布推文
When to use
- A solo dev is launching a real, usable product (not a vaporware tease).
- They have a link the reader can click (live site, Product Hunt page, GitHub repo, signup form).
- They want a single tweet or 2–4 tweet thread — not a 20-tweet manifesto.
When NOT to use
- There is no shippable artifact yet. Vague "I'm building X" tweets are noise — postpone until something is live.
- The product is enterprise / B2B with no public surface area. That belongs on LinkedIn, not X.
- The user wants a viral hook without the substance. We optimize for honest signal, not engagement-bait.
The shape of a good solo launch tweet
[Line 1: What it is, plainly]
[Line 2: Who it is for + the one job it does]
[Line 3: One concrete proof point or screenshot/video]
[Line 4: Link]
Hard rules:
- No emojis at the start. Looks like a marketer, not a builder.
- No "I'm excited to announce". Skip the throat-clearing, just announce.
- First 80 characters are everything — that's what shows in previews.
- Show, don't promise. A screenshot or 10-second clip beats adjectives.
- Mention the price if there is one (free vs $X/mo) — saves everyone time.
Output
Produce two variants:
- Single tweet (≤ 280 chars) for casual feed posts.
- Thread (3–4 tweets) for a real launch with build-up.
Put the link at the end of the single tweet, or in the last tweet of the thread (X's algorithm punishes early external links).
Worked example
User context: Launching solo-skills, a Claude Code skills pack for solo devs. Free, MIT, GitHub link.
Single tweet:
solo-skills — a Claude Code skills pack for people who ship alone.
7 skills for the things solo devs actually do daily: launch tweets, customer emails, postmortems, decision frameworks. MIT.
github.com/<you>/solo-skills
Thread:
1/ Launching solo-skills today.
A Claude Code skills pack designed for one specific person: the dev shipping a product alone.
2/ The problem with most skill packs: they assume you have a team.
solo-skills assumes you don't. Each skill is for tasks you usually do solo and badly:
- Customer emails
- Launch posts
- Postmortems
- Roadmap decisions
3/ One-line install. 7 skills.
Each comes with a "When NOT to use" section, because force-applying a framework is how solo devs waste a Saturday.
4/ MIT licensed. PRs welcome.
github.com/<you>/solo-skills
中文版
何时使用
- 单兵开发者要发布真正能用的产品(不是占坑)。
- 有可点击的链接(线上站、Product Hunt、GitHub、注册表)。
- 想要一条或3–4 条推文,不是 20 条檄文。
何时不使用
- 还没有可交付物,纯空气项目——别发,继续做。
- 纯 B2B 企业产品,没公开访问面——那种发 LinkedIn 不发 X。
- 只想要"病毒钩子"不要内容——不写营销噱头。
一条好推文的结构
[第 1 行:直白说它是什么]
[第 2 行:给谁用 + 解决什么单一问题]
[第 3 行:一个具体证据(截图/视频/数字)]
[第 4 行:链接]
硬规则:
- 开头不要 emoji——像营销不像开发者。
- 不要"很激动地宣布"——直接宣布。
- 前 80 个字符决定一切——那是预览里看到的部分。
- 展示,不要承诺——截图/10 秒视频胜过形容词。
- 明码标价——免费就写免费,付费写 $X/mo,省彼此时间。
输出
输出两个版本:
- 单推(≤ 280 字符)。
- 3–4 条线程(用于正式发布)。
链接放在单推末尾或线程最后一条(X 算法对线程里过早出现的外链不友好)。
GitHub 仓库
相关推荐技能
llamaguard
其他LlamaGuard是Meta推出的7-8B参数内容审核模型,专门用于过滤LLM的输入和输出内容。它能检测六大安全风险类别(暴力/仇恨、性内容、武器、违禁品、自残、犯罪计划),准确率达94-95%。开发者可通过HuggingFace、vLLM或Sagemaker快速部署,并能与NeMo Guardrails集成实现自动化安全防护。
cost-optimization
其他这个Claude Skill帮助开发者优化云成本,通过资源调整、标记策略和预留实例来降低AWS、Azure和GCP的开支。它适用于减少云支出、分析基础设施成本或实施成本治理策略的场景。关键功能包括提供成本可视化、资源规模调整指导和定价模型优化建议。
quantizing-models-bitsandbytes
其他这个Skill使用bitsandbytes库量化大语言模型,能在GPU内存有限时通过8位或4位量化减少50-75%内存占用,同时保持精度损失最小。它支持INT8、NF4、FP4等多种量化格式,可与HuggingFace Transformers无缝集成,适用于需要部署更大模型或加速推理的场景。还提供QLoRA训练和8位优化器支持,让开发者能轻松实现高效模型压缩。
dispatching-parallel-agents
其他该Skill用于并行处理3个以上无依赖关系的独立故障,可为每个问题域分派专属Claude代理同时执行调查修复。它通过并发处理多个独立问题显著提升故障排查效率,特别适用于测试文件、子系统等无共享状态的场景。
