
Claude Md Progressive Disclosurer
This is a methodology for taming bloated CLAUDE.md files using two-level architecture: keep high-frequency rules and code patterns in Level 1, move detailed SOPs and edge cases to Level 2 references with explicit trigger conditions. The core insight is brutal: line count is not a KPI, information loss is the enemy. It enforces "move intact, never compress" because what you think is redundant might save a future debug session. Includes a three-phase verification workflow (file existence, content completeness, no spurious deletion) and warns against fake wins like @import splitting, which doesn't actually reduce context. Written in Chinese but the principles are language-agnostic. Best used when Claude repeatedly ignores rules or you're maintaining the same information in multiple places.
- skills.sh installs
- 1K
- GitHub stars
- 1.5K
- last updated
- Aug 21, 2026
Install
$ npx -y skills add daymade/claude-code-skills --skill claude-md-progressive-disclosurer --agent claude-codeInstalls into .claude/skills of the current project.
Files
CLAUDE.md 渐进式披露优化器
核心理念
"找到最小的高信号 token 集合,最大化期望结果的可能性。" — Anthropic
目标是让指令在实际任务中被正确加载、找到并执行。 信息效率、可读性与可维护性服务于这个结果;文件大小、审阅次数和脚本绿灯都不能替代行为证据。
本 skill 在主文保留决策与执行入口,验证命令和历史材料按触发读取。官方篇幅建议用于发现可拆分内容,不是通过/失败阈值;用户需要的是正确行为,不是达到一个行数。
执行边界与验收
- 先固定当前目标文件、消费它的宿主、授权范围、用户可见结果与停止条件。诊断/审计请求只读;明确要求优化或修复时,执行范围内的本地可逆修改与必要验证。发布、不可逆操作和范围扩张按当前用户授权处理,已答过的同一事项不反复确认。
- “零信息损失”约束仍有效的契约与原样迁移,不要求旧错误永久留在现行规范里。每项改动先标为:保留/原样移动、同源去重、依据现行权威纠错,或待用户裁决的退役/边界变更。已有明确裁定按其执行;不能把“优化”当作撤销未获授权契约的许可。
- 提案或验收正文给出:当前原文 → 候选文本/准确 diff → 依据 → 行为后果 → 未验证之处。先做出可审阅的本地结果,再请求尚缺的决策;不能只贴改后版本或让用户自行翻文件拼差异。
- 依据分为当前宿主/官方契约、原始研究、用户长期契约、实测样例和待验证假设。标出研究的模型、任务、样本与限制;不能把旧模型结果、公司实践或单次成功改写成 GPT/Claude 全系通用阈值。
- 先修会改变当前任务决策的冲突、失效规则、权限歧义、假指针或真实截断。只有在问题确实是常驻负担时,才用测量贡献度安排减负顺序;不因文件最大就先改它,也不为缩小指令而新建 hook、监控或 Skill。
- 验收是:授权范围内的改动完整且无误、真实宿主加载路径正确、代表性任务符合预期。完成必要检查后停止;仅因新改动、失败或未决疑点扩大验证。普通小修改不自动加独立审阅,复杂且缺少机械裁判的改动按当前协作契约做一次有界审阅。
当前依据与适用边界见 references/progressive_disclosure_principles.md 开头;核查外部机制或研究结论时读取,历史案例不覆盖这里的现行契约。用户问「有哪些最佳实践」或要求参考外部做法时,先读那张证据表作答;是否需要重新调研,按表下的刷新规则判断。
铁律:行数禁作 KPI,可作诊断症状
禁作优化目标 / 成功指标(不可削弱——案例 7/8/9 的防线就是这条):
Sponsors
Details
- Repository
- github.com/daymade/claude-code-skills
- Categories
- Code Review & Quality, AI & Agent Building
- First seen
- Jun 3, 2026
More from daymade/claude-code-skills
All 53- Claude Code History Files Finder1K
- Github Contributor991
- Video Comparer990
- Llm Icon Finder989
- Cli Demo Generator978
- Cloudflare Troubleshooting978
- Teams Channel Post Writer970
- Claude Skills Troubleshooting960
- Repomix Safe Mixer946
- Financial Data Collector917
- Repomix Unmixer916
- Doc To Markdown914
- Product Analysis909
- Windows Remote Desktop Connection Doctor843
- Scrapling Skill827
- Skill Creator805
- Fixing Claude Export Conversations796
- Douban Skill784
- Terraform Skill746
- Marketplace Dev731
- Gangtise Copilot730
- Tunnel Doctor713
- Debugging Network Issues710
- Competitors Analysis694