返回技能列表

standup-solo

rockscy
更新于 Yesterday
2 次查看
2
1
2
在 GitHub 上查看
其他ai

关于

This skill runs a structured 5-minute personal standup for solo developers, guiding them to reflect on what they shipped yesterday, what's currently blocked, and what's next for today. It's triggered by requests for a daily check-in or when a developer needs to break out of a productivity slump. The process enforces concise, three-question reflection using the user's own notes and commits from the previous day.

快速安装

Claude Code

推荐
主要方式
npx skills add rockscy/solo-skills -a claude-code
插件命令备选方式
/plugin add https://github.com/rockscy/solo-skills
Git 克隆备选方式
git clone https://github.com/rockscy/solo-skills.git ~/.claude/skills/standup-solo

在 Claude Code 中复制并粘贴此命令以安装该技能

技能文档

Standup Solo / 单兵每日站会

When to use

  • The user is a solo operator (no team to standup with).
  • They want 5 minutes of forced reflection, not a project management session.
  • They have access to whatever notes / commits / messages they made yesterday.

When NOT to use

  • The user wants a multi-day retrospective — that's postmortem-solo.
  • They want to plan a sprint — too granular, use a planning skill instead.
  • They are clearly burned out — pushing them through standup theater is harmful. Suggest rest.

The three questions

A solo standup is three questions, hard limits:

  1. What did I actually finish yesterday? (Past, concrete, observable. "Worked on X" doesn't count — what was done?)
  2. What's blocked or about to block me? (Anything where the next step depends on someone or something outside my control.)
  3. What is the one most important thing today? (Singular. Force the user to commit.)

Output format

## Standup — YYYY-MM-DD

**Yesterday (shipped):**
- <thing 1>
- <thing 2>

**Yesterday (started but not finished):**
- <thing> — <why not done>

**Blocked:**
- <thing> — <waiting on what / who>

**Today's #1:**
> <one sentence, action-form>

**Today's #2 and #3 (only if #1 finishes):**
- <thing>
- <thing>

Inputs the user can provide

To make output sharp, the user can paste any of:

  • Yesterday's git log: git log --since="yesterday" --author=<me> --oneline
  • Sent emails / Slack DMs from yesterday
  • A screenshot of yesterday's todo app
  • Just a stream-of-consciousness brain dump

If the dev gives nothing, ask one question only: "What did you work on yesterday — give me 3 bullets, ugly is fine."

Worked example

Input (raw):

"yesterday i fixed a bug in the auth flow, started on the new pricing page but it's like 60% done, also replied to like 8 customer emails. today i need to finish pricing page and figure out why stripe webhooks are firing twice."

Output:

## Standup — 2026-04-30

**Yesterday (shipped):**
- Fixed bug in auth flow.
- Replied to 8 customer support emails.

**Yesterday (started but not finished):**
- New pricing page — ~60% done; carry over to today.

**Blocked:**
- (none reported)

**Today's #1:**
> Ship pricing page to production (the remaining 40% — copy, layout, deploy).

**Today's #2 and #3 (only if #1 finishes):**
- Investigate duplicate Stripe webhook firings.

Anti-patterns

  • Don't list hopes as accomplishments. "Worked on X" is not "shipped X". If it's not visible to a user, it didn't ship.
  • Don't have 5 #1 priorities. That's 0 priorities.
  • Don't carry blocked items silently. If a thing has been blocked 3+ days, escalate or kill it — note this in the standup.

中文版

何时使用

  • 单兵作战,没团队可以开站会。
  • 想要 5 分钟强制反省,不是项目管理会议。
  • 手头有昨天的笔记/提交/消息可参考。

何时不使用

  • 想要跨多日回顾——那是 postmortem-solo。
  • 想要冲刺规划——颗粒度不够,换规划类技能。
  • 已经明显燃尽——继续表演站会有害,建议休息。

三个问题

  1. 昨天我到底完成了什么?(过去时、具体、可观测。"在做 X"不算,要看做完了什么。)
  2. 什么阻塞了我,或快要阻塞?(下一步依赖外部人或外部物的所有事。)
  3. 今天最重要的一件事是什么?(单数。强制承诺。)

输出

## 站会 — YYYY-MM-DD

**昨日已交付:**
- <项 1>
- <项 2>

**昨日开了头没做完:**
- <项> — <为何没做完>

**阻塞:**
- <项> — <等谁/等什么>

**今日 #1:**
> <一句话,动词开头>

**今日 #2 和 #3(仅当 #1 完成后再做):**
- <项>
- <项>

反模式

  • 别把"在做"当成"做完了"——用户看不到的功能 = 没交付。
  • 别有 5 个最高优先级——那等于没优先级。
  • 别让阻塞项静默累积——超过 3 天就要升级或砍掉。

GitHub 仓库

rockscy/solo-skills
路径: skills/standup-solo
0
ai-agentsawesome-listbilingualclaude-codeclaude-skillsdeveloper-tools

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。

查看技能

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。

查看技能

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。

查看技能

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。

查看技能