作者为Codex和Claude Code中300多个Skill做上下文瘦身,发现每次新会话仅Skill列表就占约9.9k token,按7月使用强度粗算,多余Skill列表约吃掉4到5亿token的上下文空间。
After burning 500 million tokens, I came up with a new technique to slim down the Skill context for Codex and Claude Code
The author slimmed down the context for over 300 Skills in Codex and Claude Code and found that the Skill list alone takes up about 9.9k tokens in each new...
Today AI Intelligence Brief
The author slimmed down the context for over 300 Skills in Codex and Claude Code and found that the
Skill list alone takes up about 9.9k tokens in each new session. Based on the usage intensity in July, roughly, the extra Skill list consumed about 400 to 500 million tokens of context space.
Intelligence Assessment
作者针对Codex和Claude Code中的300多个Skill进行上下文瘦身,并称新会话仅加载Skill列表就约占9.9k token。
据公众号“卡尔的AI沃茨”披露,作者按7月使用强度粗略估算,多余Skill列表约消耗4亿至5亿token的上下文空间。
Aioga判断,这一案例值得关注,因为Skill数量增加后,列表本身可能成为持续的上下文开销;但材料未提供具体瘦身方法与对比结果。
对同时使用大量Skill的Codex和Claude Code用户而言,Skill列表占用可能影响可用上下文空间;实际影响仍取决于各自配置和使用强度。 建议先统计每次新会话中Skill列表的token占用,再核查低频或多余Skill;是否精简应依据实测数据,避免直接套用作者的粗略估算。
Source and Copyright
The readable text on this page was extracted from the public source and organized with attribution, publication time and the original link. Copyright remains with the original author and publisher.
Ingestion channel: Summary aggregation · Source domain: mp.weixin.qq.com
Source: 公众号:卡尔的AI沃茨
Original link: Open original source
Aioga archive: Open intelligence page
Content record: summary-fallback · Updated: 2026-08-05T06:46:00.000Z

API 中转站
统一接入主流 AI 模型 API,为开发、测试与生产环境提供稳定调用入口。
立即访问 api.w173.com