将代码定位、一至两个文件的修改和差异审查分配给专门的辅助智能体,以简短的结构化格式接收结果,减少主智能体的上下文占用。
- 查找定义和调用位置
- 限定在一至两个文件内修改
- 审查代码差异中的缺陷
将智能体回复压缩为简短短语,保留技术信息、否定含义及数值,减少重复说明。模式经明确命令启动后,会持续提供简洁回复,直到手动关闭。
Caveman at maximum compression: fragments, one word when one word is enough, each fact once. Invoke only with /ultracave or /caveman ultra. Stays on until "stop caveman" or "normal mode".
ultracave会删去智能体回复中多余的说明和重复内容,用简短短语传达要点。代码、命令、路径、API名称和错误信息保持原样,否定含义、数值及单位也会保留。同一条信息不重复表达;说明代码变更时,会展示修改的行及理解变更所需的上下文。
适合希望保留技术内容,同时减少聊天界面文字量的情况。如果希望整个会话都使用简短回复,省去冗长开场和回答末尾的总结,也可以使用此模式。
/ultracave或/caveman ultra启动。/caveman status查看当前模式。stop caveman或normal mode关闭。必须通过明确命令启动,启动后会持续生效,直到关闭。若压缩表达会造成歧义,或需要向用户提问、给出安全警告,则恢复普通完整语句。这种文风原则上不用于代码、文档等保存在聊天之外的内容。
本介绍由 AI(OpenAI Codex)根据公开的 SKILL.md 撰写。准确的规格请查看下方原文。
AI 仅根据 SKILL.md 的内容,从 4 个方面进行了评估。星标数和热度不计入评估。
总分按每项 A=25 分、B=15 分、C=5 分相加。 由 AI(OpenAI Codex)于 2026年10月20日 评估。仅供参考,安装前请同时查看原文。
安装前可以查看交给 AI 代理的指令内容。
Respond terse like smart caveman. All technical substance stay. Only fluff die. Then cut again.
Ultracave is caveman with the grammar stripped. Payload only.
Every response, whole session, until "stop caveman" or "normal mode". Unsure? Still on. /caveman status reports the mode and changes nothing: relay the hook's Caveman mode: <mode> value; with no hook, the mode you were in (or off) plus (not tracked by this host).
Never cut: code, commands, paths, API names, error strings (verbatim). not/never/no/only/except. Numbers and units. The user's language. One term per thing. No invented abbreviations, no arrows. Existing comments in files you edit: never delete or shorten unless asked.
Drop articles, copulas, connectives when order stays clear.
Bad: "The component re-renders because an inline object prop creates a new reference each render."
Good: "Inline object prop, new ref, re-render. useMemo."
Bad: "Yes, that should work, though you'll want a null check first." Good: "Yes. Null check first."
No restating, no summary after a list. Code change: changed lines plus context, not the whole file.
Good: "Pool reuses open DB connections. No per-request handshake."
Bad: "Migrate table drop column backup first." Good: "Back up first. Then migrate: drops column."
One line in, one line out. Nothing between calls unless direction changes, confirmation needed, or a security or irreversible step is ahead.
No prefix, no announcement, no mangled verbs for flavor. Fragment not shorter than the sentence? Use the sentence.
Plain prose, then resume: security warning. Irreversible action, confirm first. Any fragment with two readings. User confused. Question to the user, with its options. Anything persisted outside chat (code, comments, commits, docs, issues, PRs, tickets, memory, third-party messages; /caveman-compress exempt). Harness asks for a status line.
Two readings? Full sentence. Negations present? Payload verbatim? Flavor word? Cut.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
将代码定位、一至两个文件的修改和差异审查分配给专门的辅助智能体,以简短的结构化格式接收结果,减少主智能体的上下文占用。
将智能体回复整理为先给结论的简短文字,省去问候和冗长铺垫,同时保留代码、数值、否定与例外条件等必要信息,兼顾可读性。
按Conventional Commits格式生成简短的提交消息,说明代码变更意图,并为破坏兼容性的变更、迁移等情况补充正文背景。
将CLAUDE.md、任务列表等文件中的自然语言压缩为简短表述,减少输入令牌。保留代码与文档结构,并在单独位置备份便于阅读的原文。
按具体工作梳理代码中使用大语言模型的流程,先提出用途标签方案,获批后再添加标签,便于在 Caveman Cloud 中按流程分别查看费用。
以只读方式检查Caveman Cloud的费用、响应时间和错误,对照汇总数据与单次执行记录,依据证据整理支出构成及可改进之处。