菜单
Preserve output from expensive one-shot commands for monitoring and later analysis. Use when rerunning a build, API query, CI inspection, or diagnostic would be costly and a managed background process is not the better interface.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/aspiers/ai-config --skill slow-command-running命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | slow-command-running |
| description | Preserve output from expensive one-shot commands for monitoring and later analysis. Use when rerunning a build, API query, CI inspection, or diagnostic would be costly and a managed background process is not the better interface. |
Capture output when it will be useful during or after a slow command. For servers, watchers, and ongoing log streams, prefer the harness's managed background-process tool instead of a shell pipeline.
Choose a repository-approved ignored path, then preserve both displayed and saved output:
mkdir -p tmp
command 2>&1 | tee tmp/descriptive-command.log
Use set -o pipefail or inspect the producer's pipeline status when command
success matters; tee succeeding does not prove the original command
succeeded.
Name logs by purpose and stable identifier so concurrent runs do not overwrite one another. Avoid recording credentials or sensitive response data.
Do not add tee mechanically to fast commands or output that will be consumed
only once. Avoid truncating a live producer with head unless early SIGPIPE
termination is intended.