[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"extension-skill-juliusbrussee-caveman-compress-zh-CN":3,"guides-for-juliusbrussee-caveman-compress":503,"similar-k17adsjvvcgs7b6d0p7qkjv6js86mryx-zh-CN":504},{"_creationTime":4,"_id":5,"children":6,"community":7,"display":9,"evaluation":15,"identity":242,"isFallback":224,"parentExtension":247,"providers":303,"relations":307,"repo":309,"tags":500,"workflow":501},1778692122841.1462,"k17adsjvvcgs7b6d0p7qkjv6js86mryx",[],{"reviewCount":8},0,{"description":10,"installMethods":11,"name":13,"sourceUrl":14},"将自然语言内存文件（CLAUDE.md、todos、preferences）压缩为 caveman 格式，以节省输入 token。保留所有技术内容、代码、URL 和结构。压缩版本会覆盖原始文件。人类可读的备份保存为 FILE.original.md。触发方式：/caveman-compress FILEPATH 或 \"compress memory file\"\n",{"claudeCode":12},"juliusbrussee/caveman","caveman-compress","https://github.com/juliusbrussee/caveman",{"_creationTime":16,"_id":17,"extensionId":5,"locale":18,"result":19,"trustSignals":222,"workflow":240},1778692122841.1465,"kn774getg0a6chtf0s4wteqbph86ndv3","zh-CN",{"checks":20,"evaluatedAt":192,"extensionSummary":193,"features":194,"nonGoals":200,"promptVersionExtension":205,"promptVersionScoring":206,"purpose":207,"rationale":208,"score":209,"summary":210,"tags":211,"tier":217,"useCases":218},[21,26,29,32,36,39,44,48,51,54,58,62,65,69,72,75,78,81,84,87,91,95,99,103,107,110,113,116,120,123,126,129,132,135,138,142,146,150,153,157,160,163,166,169,173,176,179,182,185,189],{"category":22,"check":23,"severity":24,"summary":25},"Practical Utility","Problem relevance","pass","描述清楚地指出了通过压缩内存文件来节省输入 token 的问题，并指出了具体的产物（.md、todos、preferences）和用户意图（节省 token、降低成本）。",{"category":22,"check":27,"severity":24,"summary":28},"Unique selling proposition","该技能通过为自然语言文件提供专门的压缩算法，在保留技术内容的同时减少 token 数量，超越了简单的摘要或改写，提供了比默认 LLM 行为显著的价值。",{"category":22,"check":30,"severity":24,"summary":31},"Production readiness","该技能已为生产做好准备，涵盖了文件压缩的整个生命周期：它检测文件类型、压缩、验证、处理错误并重试、备份原始文件，并覆盖压缩后的文件，所有这些都经过测试。",{"category":33,"check":34,"severity":24,"summary":35},"Scope","Single responsibility principle","该技能仅专注于将自然语言文件压缩为更节省 token 的格式，其边界清晰，明确了它将压缩什么以及不压缩什么。",{"category":33,"check":37,"severity":24,"summary":38},"Description quality","显示的描述准确地反映了技能的功能，包括其触发方式、目的、备份机制和文件保留规则。",{"category":40,"check":41,"severity":42,"summary":43},"Invocation","Scoped tools","not_applicable","该技能是一个单一的命令行工具，不暴露多个可供评估范围的独立工具。",{"category":45,"check":46,"severity":42,"summary":47},"Documentation","Configuration & parameter reference","除了 Anthropic API 密钥（由环境或 CLI 配置处理）之外，该技能没有可配置的参数或环境变量。",{"category":33,"check":49,"severity":24,"summary":50},"Tool naming","该技能的主要命令是 `/caveman-compress`，它具有描述性。",{"category":33,"check":52,"severity":24,"summary":53},"Minimal I/O surface","输入是单个文件路径，输出是修改后的文件和备份，符合其任务所需的最小 I/O。",{"category":55,"check":56,"severity":24,"summary":57},"License","License usability","该扩展根据 MIT 许可证分发，如捆绑的 LICENSE 文件所示，这是一种宽松的开源许可证。",{"category":59,"check":60,"severity":24,"summary":61},"Maintenance","Commit recency","上次提交是在 2026 年 5 月 12 日，在最近 3 个月内。",{"category":59,"check":63,"severity":24,"summary":64},"Dependency Management","该技能使用 Python，并依赖于 `anthropic` 库和标准的 Python 库。Python 的打包和环境管理足以满足此技能的需求。",{"category":66,"check":67,"severity":24,"summary":68},"Security","Secret Management","该技能通过环境变量或 CLI 身份验证处理 API 密钥，并且其 SECURITY.md 明确表示不回显密钥并拒绝压缩敏感文件类型。",{"category":66,"check":70,"severity":24,"summary":71},"Injection","该技能旨在将文件内容视为数据，并明确避免将其作为代码执行。它通过拒绝处理敏感路径来处理敏感路径。",{"category":66,"check":73,"severity":24,"summary":74},"Transitive Supply-Chain Grenades","该技能处理本地文件并直接或通过 CLI 调用 Anthropic API，没有运行时获取外部代码或指令。",{"category":66,"check":76,"severity":24,"summary":77},"Sandbox Isolation","该技能仅操作指定的文件并写入同一位置，遵守沙箱边界。",{"category":66,"check":79,"severity":24,"summary":80},"Sandbox escape primitives","Python 脚本使用标准库调用和 `subprocess.run` 而不带 shell=True，避免了常见的沙箱逃逸原语。",{"category":66,"check":82,"severity":24,"summary":83},"Data Exfiltration","该技能旨在通过拒绝处理敏感文件类型并仅与 Anthropic API 交互进行压缩来避免泄露机密数据。",{"category":66,"check":85,"severity":24,"summary":86},"Hidden Text Tricks","捆绑的脚本和 markdown 文件不包含隐藏文本技巧或混淆内容。",{"category":88,"check":89,"severity":24,"summary":90},"Hooks","Opaque code execution","Python 脚本以清晰、可读的源代码形式提供，没有混淆、base64 编码或运行时下载。",{"category":92,"check":93,"severity":24,"summary":94},"Portability","Structural Assumption","该脚本操作用户提供的文件路径，不假设项目目录结构。",{"category":96,"check":97,"severity":24,"summary":98},"Trust","Issues Attention","在过去 90 天内，共打开 66 个问题和关闭 70 个问题，关闭率约为 51%，表明维护者参与度良好。",{"category":100,"check":101,"severity":24,"summary":102},"Versioning","Release Management","文件 `scripts/__init__.py` 声明了 `__version__ = \"1.0.0\"`，表明脚本包有一个可检测的版本。",{"category":104,"check":105,"severity":24,"summary":106},"Code Execution","Validation","验证脚本检查标题、代码块、URL、路径和内联代码的精确保留，确保完整性。",{"category":66,"check":108,"severity":24,"summary":109},"Unguarded Destructive Operations","该技能会覆盖原始文件，但会先创建备份，并且除了这种受控修改之外，不执行任何破坏性操作。",{"category":104,"check":111,"severity":24,"summary":112},"Error Handling","脚本 `compress.py` 包括重试、检查 Claude 失败、空响应、相同输出，并在验证失败时恢复原始文件，并清楚地报告错误。",{"category":104,"check":114,"severity":24,"summary":115},"Logging","脚本为处理步骤、验证尝试以及成功/失败消息提供控制台输出，作为基本审计日志。",{"category":117,"check":118,"severity":24,"summary":119},"Compliance","GDPR","该技能拒绝压缩带有敏感名称或路径的文件，防止意外将 PII 或凭据提交给 Anthropic API。",{"category":117,"check":121,"severity":24,"summary":122},"Target market","该技能操作本地文件并与 Anthropic API 交互，没有区域或司法管辖区限制；目标市场是全球性的。",{"category":92,"check":124,"severity":24,"summary":125},"Runtime stability","脚本指定 Python 3.10+ 并使用标准库，确保广泛的 POSIX 兼容性，并在 SDK 不可用时优雅地回退到 CLI。",{"category":45,"check":127,"severity":24,"summary":128},"README","README 提供了清晰的概述、安装说明、基准测试和使用示例，有效地陈述了扩展的目的。",{"category":33,"check":130,"severity":42,"summary":131},"Tool surface size","这是一个通过命令行界面调用的单一用途技能，而不是一组工具。",{"category":40,"check":133,"severity":42,"summary":134},"Overlapping near-synonym tools","该技能只公开一个主要命令，因此没有重叠的工具。",{"category":45,"check":136,"severity":24,"summary":137},"Phantom features","所有宣传的功能，如压缩特定文件类型和保留规则，都在代码和文档中得到实现和验证。",{"category":139,"check":140,"severity":24,"summary":141},"Install","Installation instruction","README 提供了 `caveman` 插件的清晰安装说明和 `/caveman-compress` 命令的使用示例。",{"category":143,"check":144,"severity":24,"summary":145},"Errors","Actionable error messages","错误会报告上下文（例如，文件未找到、文件过大、敏感路径、验证失败），并且通常包含补救步骤或明确的失败原因。",{"category":147,"check":148,"severity":24,"summary":149},"Execution","Pinned dependencies","脚本指定 `python3` 和 `#!/usr/bin/env python3`，并且虽然在 lock 文件中没有固定特定的库版本，但使用标准 Python 库使其具有可移植性。",{"category":33,"check":151,"severity":42,"summary":152},"Dry-run preview","该技能的核心操作是文件修改。预览模式不适用，因为主要操作是压缩和覆盖。",{"category":154,"check":155,"severity":42,"summary":156},"Protocol","Idempotent retry & timeouts","该操作是针对本地文件的，而不是远程调用，并且虽然为 LLM 调用实现了重试，但没有为本地操作显式配置超时。",{"category":117,"check":158,"severity":24,"summary":159},"Telemetry opt-in","该技能不发送任何遥测数据，默认满足选择加入要求。",{"category":40,"check":161,"severity":24,"summary":162},"Precise Purpose","该技能清楚地说明它将自然语言文件（CLAUDE.md、todos、preferences）压缩为 caveman 格式以节省输入 token，并具有特定的触发方式和明确的边界。",{"category":40,"check":164,"severity":24,"summary":165},"Concise Frontmatter","SKILL.md 中的 frontmatter 简洁明了，在合理的字符限制内清楚地说明了名称、描述、触发方式和目的。",{"category":45,"check":167,"severity":24,"summary":168},"Concise Body","SKILL.md 的主体内容简洁，概述了过程和规则，并将详细解释委托给单独的文件或内联文档。",{"category":170,"check":171,"severity":24,"summary":172},"Context","Progressive Disclosure","SKILL.md 简洁明了，虽然没有明确链接到单独的 `reference/` 文件，但它在内联提供了足够多的细节，而没有嵌入大块内容。",{"category":170,"check":174,"severity":42,"summary":175},"Forked exploration","该技能是直接的文件操作工具，不涉及需要 fork 上下文的深度探索或代码审查。",{"category":22,"check":177,"severity":24,"summary":178},"Usage examples","README 提供了清晰的使用示例（包含文件路径），SKILL.md 详细说明了命令结构，演示了如何调用该技能。",{"category":22,"check":180,"severity":24,"summary":181},"Edge cases","该技能记录并处理了边缘情况，如空文件、大文件、敏感文件名、相同输出和验证失败，并记录了恢复步骤。",{"category":104,"check":183,"severity":24,"summary":184},"Tool Fallback","该脚本如果设置了 API 密钥，则使用 `anthropic` SDK，如果未设置，则回退到 `claude` CLI 进行身份验证，充当功能回退。",{"category":186,"check":187,"severity":24,"summary":188},"Safety","Halt on unexpected state","脚本在遇到意外状态（如空文件、大文件、敏感路径或 Claude 返回无法使用的输出）时停止压缩，并在持续验证失败时恢复原始文件。",{"category":92,"check":190,"severity":24,"summary":191},"Cross-skill coupling","该技能是独立的，不隐式依赖于同一会话中的其他已加载技能。",1778691883697,"此技能将 .md、.txt 和 .tex 等自然语言文件压缩为更节省 token 的“caveman”格式。它保留代码块、URL 和结构，覆盖原始文件，并保存人类可读的备份。它使用 Anthropic SDK 或 CLI，并包含强大的验证和错误处理。",[195,196,197,198,199],"压缩自然语言文件以节省 token","精确保留代码块、URL 和文件路径","用压缩版本覆盖原始文件","保存原始文件的人类可读备份","处理验证、重试和错误恢复",[201,202,203,204],"压缩代码文件（.py、.js、.json 等）","压缩配置文件（.yaml、.env、.toml）","修改或删除原始文件而不创建备份","压缩敏感文件，如凭证或私钥","3.0.0","4.4.0","通过将自然语言内存文件压缩为更简洁的格式来节省 Claude 输入 token 并降低会话成本。","该扩展在代码质量、健壮的错误处理和全面的文档方面表现出色。由于在“依赖项管理”检查中只有一个“pass”，这对于其使用标准库不适用，因此其得分上限非常低。",99,"一个高质量、生产就绪的技能，用于压缩自然语言文件以节省 Claude token。",[212,213,214,215,216],"compression","token-saving","markdown","text-processing","utility","verified",[219,220,221],"通过压缩大型 CLAUDE.md 或偏好文件来降低 Claude 会话成本","使项目笔记和待办事项列表对频繁加载更节省 token","在不丢失技术内容或代码的情况下节省输入 token",{"codeQuality":223,"collectedAt":225,"documentation":226,"maintenance":229,"popularity":235,"security":236,"testCoverage":239},{"hasLockfile":224},false,1778691864302,{"descriptionLength":227,"readmeSize":228},335,10896,{"closedIssues90d":230,"forks":231,"hasChangelog":224,"openIssues90d":232,"pushedAt":233,"stars":234},70,3292,66,1778615088000,59509,{"npmDownloads":8},{"hasNpmPackage":237,"license":238,"smitheryVerified":224},true,"MIT",{"hasCi":237,"hasTests":237},{"updatedAt":241},1778692122841,{"basePath":243,"githubOwner":244,"githubRepo":245,"locale":18,"slug":13,"type":246},"skills/caveman-compress","juliusbrussee","caveman","skill",{"_creationTime":248,"_id":249,"community":250,"display":251,"identity":255,"parentExtension":259,"providers":291,"relations":298,"tags":299,"workflow":300},1778691765524.373,"k170g496f5ze5hv8xa2e7b91ex86mz9m",{"reviewCount":8},{"description":252,"installMethods":253,"name":254,"sourceUrl":14},"Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.",{"claudeCode":245},"Caveman",{"basePath":256,"githubOwner":244,"githubRepo":245,"locale":257,"slug":245,"type":258},"","en","plugin",{"_creationTime":260,"_id":261,"community":262,"display":263,"identity":266,"providers":268,"relations":284,"tags":286,"workflow":287},1778691765524.3728,"k17dcfwpwpzjc43zt2qawvft3h86mmse",{"reviewCount":8},{"description":264,"installMethods":265,"name":245,"sourceUrl":14},"Ultra-compressed communication mode for Claude Code. Cuts ~75% of tokens while keeping full technical accuracy.",{"claudeCode":12},{"basePath":256,"githubOwner":244,"githubRepo":245,"locale":257,"slug":245,"type":267},"marketplace",{"evaluate":269,"extract":278},{"promptVersionExtension":270,"promptVersionScoring":206,"score":271,"tags":272,"targetMarket":277,"tier":217},"3.1.0",92,[212,273,274,275,276],"llm","productivity","token-reduction","developer-tools","global",{"commitSha":279,"marketplace":280,"plugin":282},"HEAD",{"name":245,"pluginCount":281},1,{"mcpCount":8,"provider":283,"skillCount":8},"classify",{"repoId":285},"kd7890f0sfy0ymf1pakpx7vqch86nmew",[212,276,273,274,275],{"evaluatedAt":288,"extractAt":289,"updatedAt":290},1778691801317,1778691765524,1778692173865,{"evaluate":292,"extract":295},{"promptVersionExtension":205,"promptVersionScoring":206,"score":209,"tags":293,"targetMarket":277,"tier":217},[212,273,294,276,274],"text",{"commitSha":279,"license":238,"plugin":296},{"mcpCount":8,"provider":283,"skillCount":297},6,{"parentExtensionId":261,"repoId":285},[212,276,273,274,294],{"evaluatedAt":301,"extractAt":289,"updatedAt":302},1778691819811,1778692174193,{"evaluate":304,"extract":306},{"promptVersionExtension":205,"promptVersionScoring":206,"score":209,"tags":305,"targetMarket":277,"tier":217},[212,213,214,215,216],{"commitSha":279},{"parentExtensionId":249,"repoId":285,"translatedFrom":308},"k176a757cfd35jb9npje9qqzed86nvpn",{"_creationTime":310,"_id":285,"identity":311,"providers":312,"workflow":495},1778691760140.7458,{"githubOwner":244,"githubRepo":245,"sourceUrl":14},{"classify":313,"discover":479,"extract":482,"github":483,"npm":494},{"commitSha":279,"extensions":314},[315,325,336,380,389,397,420,429,438,447,456,468],{"basePath":316,"displayName":317,"installMethods":318,"rationale":319,"selectedPaths":320,"source":324,"sourceLanguage":257,"type":267},".agents/plugins","caveman-repo",{"claudeCode":12},"marketplace.json at .agents/plugins/marketplace.json",[321],{"path":322,"priority":323},"marketplace.json","mandatory","rule",{"basePath":256,"description":264,"displayName":245,"installMethods":326,"rationale":327,"selectedPaths":328,"source":324,"sourceLanguage":257,"type":267},{"claudeCode":12},"marketplace.json at .claude-plugin/marketplace.json",[329,331,333],{"path":330,"priority":323},".claude-plugin/marketplace.json",{"path":332,"priority":323},"README.md",{"path":334,"priority":335},"LICENSE","high",{"basePath":256,"description":252,"displayName":245,"installMethods":337,"rationale":338,"selectedPaths":339,"source":324,"sourceLanguage":257,"type":258},{"claudeCode":245},"plugin manifest at .claude-plugin/plugin.json (coalesced with duplicate plugin at plugins/caveman)",[340,342,343,344,347,349,351,353,355,357,359,361,363,365,368,370,372,374,376,378],{"path":341,"priority":323},".claude-plugin/plugin.json",{"path":332,"priority":323},{"path":334,"priority":335},{"path":345,"priority":346},"skills/cavecrew/SKILL.md","medium",{"path":348,"priority":346},"skills/caveman/SKILL.md",{"path":350,"priority":346},"skills/caveman-commit/SKILL.md",{"path":352,"priority":346},"skills/caveman-compress/SKILL.md",{"path":354,"priority":346},"skills/caveman-help/SKILL.md",{"path":356,"priority":346},"skills/caveman-review/SKILL.md",{"path":358,"priority":346},"skills/caveman-stats/SKILL.md",{"path":360,"priority":335},"agents/cavecrew-builder.md",{"path":362,"priority":335},"agents/cavecrew-investigator.md",{"path":364,"priority":335},"agents/cavecrew-reviewer.md",{"path":366,"priority":367},"bin/install.js","low",{"path":369,"priority":367},"bin/lib/openclaw.js",{"path":371,"priority":367},"bin/lib/settings.js",{"path":373,"priority":335},"commands/caveman-commit.toml",{"path":375,"priority":335},"commands/caveman-init.toml",{"path":377,"priority":335},"commands/caveman-review.toml",{"path":379,"priority":335},"commands/caveman.toml",{"basePath":381,"description":382,"displayName":383,"installMethods":384,"rationale":385,"selectedPaths":386,"source":324,"sourceLanguage":257,"type":246},".junie/skills/cavecrew","Decision guide for delegating to caveman-style subagents. Tells the main thread WHEN to spawn `cavecrew-investigator` (locate code), `cavecrew-builder` (1-2 file edit), or `cavecrew-reviewer` (diff review) instead of doing the work inline or using vanilla `Explore`. Subagent output is caveman-compressed so the tool-result injected back into main context is ~60% smaller — main context lasts longer across long sessions. Trigger: \"delegate to subagent\", \"use cavecrew\", \"spawn investigator/builder/reviewer\", \"save context\", \"compressed agent output\".\n","cavecrew",{"claudeCode":12},"SKILL.md frontmatter at .junie/skills/cavecrew/SKILL.md (coalesced with duplicate skill at .agents/skills/cavecrew, skills/cavecrew, .roo/skills/cavecrew, .kiro/skills/cavecrew, plugins/caveman/skills/cavecrew)",[387],{"path":388,"priority":323},"SKILL.md",{"basePath":390,"description":391,"displayName":245,"installMethods":392,"rationale":393,"selectedPaths":394,"source":324,"sourceLanguage":257,"type":246},"skills/caveman","Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says \"caveman mode\", \"talk like caveman\", \"use caveman\", \"less tokens\", \"be brief\", or invokes /caveman. Also auto-triggers when token efficiency is requested.\n",{"claudeCode":12},"SKILL.md frontmatter at skills/caveman/SKILL.md (coalesced with duplicate skill at plugins/caveman/skills/caveman)",[395,396],{"path":388,"priority":323},{"path":332,"priority":335},{"basePath":243,"description":398,"displayName":13,"installMethods":399,"rationale":400,"selectedPaths":401,"source":324,"sourceLanguage":257,"type":246},"Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman-compress FILEPATH or \"compress memory file\"\n",{"claudeCode":12},"SKILL.md frontmatter at skills/caveman-compress/SKILL.md (coalesced with duplicate skill at plugins/caveman/skills/caveman-compress)",[402,403,404,406,408,410,412,414,416,418],{"path":388,"priority":323},{"path":332,"priority":335},{"path":405,"priority":346},"SECURITY.md",{"path":407,"priority":367},"scripts/__init__.py",{"path":409,"priority":367},"scripts/__main__.py",{"path":411,"priority":367},"scripts/benchmark.py",{"path":413,"priority":367},"scripts/cli.py",{"path":415,"priority":367},"scripts/compress.py",{"path":417,"priority":367},"scripts/detect.py",{"path":419,"priority":367},"scripts/validate.py",{"basePath":421,"description":422,"displayName":423,"installMethods":424,"rationale":425,"selectedPaths":426,"source":324,"sourceLanguage":257,"type":246},"skills/caveman-stats","Show real token usage and estimated savings for the current session. Reads directly from the Claude Code session log — no AI estimation. Triggers on /caveman-stats. Output is injected by the mode-tracker hook; the model itself does not compute the numbers.\n","caveman-stats",{"claudeCode":12},"SKILL.md frontmatter at skills/caveman-stats/SKILL.md (coalesced with duplicate skill at plugins/caveman/skills/caveman-stats)",[427,428],{"path":388,"priority":323},{"path":332,"priority":335},{"basePath":430,"description":431,"displayName":432,"installMethods":433,"rationale":434,"selectedPaths":435,"source":324,"sourceLanguage":257,"type":246},"skills/caveman-commit","Ultra-compressed commit message generator. Cuts noise from commit messages while preserving intent and reasoning. Conventional Commits format. Subject ≤50 chars, body only when \"why\" isn't obvious. Use when user says \"write a commit\", \"commit message\", \"generate commit\", \"/commit\", or invokes /caveman-commit. Auto-triggers when staging changes.\n","caveman-commit",{"claudeCode":12},"SKILL.md frontmatter at skills/caveman-commit/SKILL.md",[436,437],{"path":388,"priority":323},{"path":332,"priority":335},{"basePath":439,"description":440,"displayName":441,"installMethods":442,"rationale":443,"selectedPaths":444,"source":324,"sourceLanguage":257,"type":246},"skills/caveman-help","Quick-reference card for all caveman modes, skills, and commands. One-shot display, not a persistent mode. Trigger: /caveman-help, \"caveman help\", \"what caveman commands\", \"how do I use caveman\".\n","caveman-help",{"claudeCode":12},"SKILL.md frontmatter at skills/caveman-help/SKILL.md",[445,446],{"path":388,"priority":323},{"path":332,"priority":335},{"basePath":448,"description":449,"displayName":450,"installMethods":451,"rationale":452,"selectedPaths":453,"source":324,"sourceLanguage":257,"type":246},"skills/caveman-review","Ultra-compressed code review comments. Cuts noise from PR feedback while preserving the actionable signal. Each comment is one line: location, problem, fix. Use when user says \"review this PR\", \"code review\", \"review the diff\", \"/review\", or invokes /caveman-review. Auto-triggers when reviewing pull requests.\n","caveman-review",{"claudeCode":12},"SKILL.md frontmatter at skills/caveman-review/SKILL.md",[454,455],{"path":388,"priority":323},{"path":332,"priority":335},{"basePath":256,"description":457,"displayName":458,"installMethods":459,"license":238,"rationale":460,"selectedPaths":461,"source":324,"sourceLanguage":257,"type":467},"Caveman installer — detects your AI coding agents and installs caveman for each one.","caveman-installer",{"npm":458},"cli ecosystem detected at /",[462,464,465,466],{"path":463,"priority":323},"package.json",{"path":332,"priority":323},{"path":334,"priority":335},{"path":366,"priority":346},"cli",{"basePath":469,"description":470,"displayName":471,"installMethods":472,"license":238,"rationale":473,"selectedPaths":474,"source":324,"sourceLanguage":257,"type":467},"src/mcp-servers/caveman-shrink","MCP proxy that compresses prose fields (tool descriptions, etc.) using caveman rules. Same accuracy, fewer context tokens.","caveman-shrink",{"npm":471},"cli ecosystem detected at src/mcp-servers/caveman-shrink",[475,476,477],{"path":463,"priority":323},{"path":332,"priority":323},{"path":478,"priority":346},"index.js",{"sources":480},[481],"manual",{"npmPackage":458},{"closedIssues90d":230,"description":484,"forks":231,"homepage":485,"license":238,"openIssues90d":232,"pushedAt":233,"readmeSize":228,"stars":234,"topics":486},"🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman","https://getcaveman.dev/",[487,488,245,489,490,273,491,492,246,493],"ai","anthropic","claude","claude-code","meme","prompt-engineering","tokens",{"downloads":8},{"classifiedAt":496,"discoverAt":497,"extractAt":498,"githubAt":498,"npmAt":499,"updatedAt":496},1778691765329,1778691760140,1778691762475,1778691764072,[212,214,215,213,216],{"evaluatedAt":502,"extractAt":289,"updatedAt":241},1778691883875,[],[505,533,564,593,622,651],{"_creationTime":506,"_id":507,"community":508,"display":509,"identity":515,"providers":519,"relations":526,"tags":529,"workflow":530},1778692846401.1528,"k17ah0hhhwvmsj6h6vxmfmd9nn86mhff",{"reviewCount":8},{"description":510,"installMethods":511,"name":513,"sourceUrl":514},"Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Use when user says \"caveman mode\", \"talk like caveman\", \"use caveman\", \"less tokens\", \"be brief\", or invokes /caveman.\n",{"claudeCode":512},"mattpocock/skills","Caveman Mode","https://github.com/mattpocock/skills",{"basePath":516,"githubOwner":517,"githubRepo":518,"locale":257,"slug":245,"type":246},"skills/productivity/caveman","mattpocock","skills",{"evaluate":520,"extract":525},{"promptVersionExtension":205,"promptVersionScoring":206,"score":521,"tags":522,"targetMarket":277,"tier":217},98,[523,212,524,273,294,216],"communication","efficiency",{"commitSha":279,"license":238},{"parentExtensionId":527,"repoId":528},"k171mc92f05dc4v6jzgf6bng7d86mr9t","kd70nm2vazehgqc4j28ve6cwv986mv1s",[523,212,524,273,294,216],{"evaluatedAt":531,"extractAt":532,"updatedAt":531},1778693459372,1778692846401,{"_creationTime":534,"_id":535,"community":536,"display":537,"identity":543,"providers":548,"relations":557,"tags":560,"workflow":561},1778694269038.6682,"k1752cypc448mke749yjbkc65186mg6f",{"reviewCount":8},{"description":538,"installMethods":539,"name":541,"sourceUrl":542},"This skill should be used when the user asks to \"compress context\", \"summarize conversation history\", \"implement compaction\", \"reduce token usage\", or mentions context compression, structured summarization, tokens-per-task optimization, or long-running agent sessions exceeding context limits.",{"claudeCode":540},"muratcankoylan/Agent-Skills-for-Context-Engineering","Context Compression","https://github.com/muratcankoylan/Agent-Skills-for-Context-Engineering",{"basePath":544,"githubOwner":545,"githubRepo":546,"locale":257,"slug":547,"type":246},"skills/context-compression","muratcankoylan","Agent-Skills-for-Context-Engineering","context-compression",{"evaluate":549,"extract":556},{"promptVersionExtension":205,"promptVersionScoring":206,"score":550,"tags":551,"targetMarket":277,"tier":217},100,[552,273,553,554,212,555],"context-engineering","agent","summarization","evaluation",{"commitSha":279,"license":238},{"parentExtensionId":558,"repoId":559},"k1754dy3wbsv2a5gr1a983zzs586njca","kd7f12maf5nxmx5xttjx7scfnx86m1tv",[553,212,552,555,273,554],{"evaluatedAt":562,"extractAt":563,"updatedAt":562},1778694410149,1778694269038,{"_creationTime":565,"_id":566,"community":567,"display":568,"identity":574,"providers":578,"relations":586,"tags":589,"workflow":590},1778691497595.9666,"k170aa7za6n5wvxtwx7xj8dc9d86mx9e",{"reviewCount":8},{"description":569,"installMethods":570,"name":572,"sourceUrl":573},"Compresses images to WebP (default) or PNG with automatic tool selection. Use when user asks to \"compress image\", \"optimize image\", \"convert to webp\", or reduce image file size.",{"claudeCode":571},"jimliu/baoyu-skills","baoyu-compress-image","https://github.com/jimliu/baoyu-skills",{"basePath":575,"githubOwner":576,"githubRepo":577,"locale":257,"slug":572,"type":246},"skills/baoyu-compress-image","jimliu","baoyu-skills",{"evaluate":579,"extract":585},{"promptVersionExtension":205,"promptVersionScoring":206,"score":550,"tags":580,"targetMarket":277,"tier":217},[581,212,582,583,584],"image","webp","png","optimization",{"commitSha":279},{"parentExtensionId":587,"repoId":588},"k172q4hxm2acmg5ree5aeemx1x86m6hx","kd718jn448p61w289hzf15tecd86n3a4",[212,581,584,583,582],{"evaluatedAt":591,"extractAt":592,"updatedAt":591},1778691619519,1778691497596,{"_creationTime":594,"_id":595,"community":596,"display":597,"identity":603,"providers":606,"relations":614,"tags":617,"workflow":618},1778699327207.9,"k175rvcd9dmnjemnr3t64br1vh86nx2b",{"reviewCount":8},{"description":598,"installMethods":599,"name":601,"sourceUrl":602},"当用户想要翻译存储库 README、使存储库支持多语言、本地化文档、添加语言切换器、国际化 README 或更新 GitHub 风格存储库中的本地化 README 变体时使用。",{"claudeCode":600},"xixu-me/skills","readme-i18n","https://github.com/xixu-me/skills",{"basePath":604,"githubOwner":605,"githubRepo":518,"locale":18,"slug":601,"type":246},"skills/readme-i18n","xixu-me",{"evaluate":607,"extract":613},{"promptVersionExtension":205,"promptVersionScoring":206,"score":550,"tags":608,"targetMarket":277,"tier":217},[609,610,611,214,612],"localization","documentation","internationalization","github",{"commitSha":279},{"repoId":615,"translatedFrom":616},"kd77r2vb42jmgam0qbr9f2c6kn86mebv","k173sze6h2kdjhyfbveynf98t586n5wt",[610,612,611,609,214],{"evaluatedAt":619,"extractAt":620,"updatedAt":621},1778699174967,1778699106670,1778699327208,{"_creationTime":623,"_id":624,"community":625,"display":626,"identity":632,"providers":636,"relations":644,"tags":647,"workflow":648},1778691104676.0076,"k17bs66raa8vz3q5ysjnzhtvc586m7j3",{"reviewCount":8},{"description":627,"installMethods":628,"name":630,"sourceUrl":631},"Generate CSV, Markdown, and XLSX spreadsheets from structured tabular data.",{"claudeCode":629},"iterationlayer/skills","Spreadsheet Generation API","https://github.com/iterationlayer/skills",{"basePath":633,"githubOwner":634,"githubRepo":518,"locale":257,"slug":635,"type":246},"skills/sheet-generation-api","iterationlayer","sheet-generation-api",{"evaluate":637,"extract":643},{"promptVersionExtension":205,"promptVersionScoring":206,"score":550,"tags":638,"targetMarket":277,"tier":217},[639,640,641,214,642],"spreadsheet","csv","xlsx","data-generation",{"commitSha":279,"license":238},{"parentExtensionId":645,"repoId":646},"k1721s0xmp59902ybtpakrrffn86n10s","kd76p4g2qmtrkgx99cnab3683d86n4g8",[640,642,214,639,641],{"evaluatedAt":649,"extractAt":650,"updatedAt":649},1778693905213,1778691104676,{"_creationTime":652,"_id":653,"community":654,"display":655,"identity":661,"providers":666,"relations":674,"tags":676,"workflow":677},1778691799740.4802,"k173gtrv9a9ev95755pa3w820h86n7r2",{"reviewCount":8},{"description":656,"installMethods":657,"name":659,"sourceUrl":660},"Comprehensive markdown and Mermaid diagram writing skill. Use when creating any scientific document, report, analysis, or visualization. Establishes text-based diagrams as the default documentation standard with full style guides (markdown + mermaid), 24 diagram type references, and 9 document templates.",{"claudeCode":658},"K-Dense-AI/claude-scientific-skills","Markdown and Mermaid Writing","https://github.com/K-Dense-AI/claude-scientific-skills",{"basePath":662,"githubOwner":663,"githubRepo":664,"locale":257,"slug":665,"type":246},"scientific-skills/markdown-mermaid-writing","K-Dense-AI","claude-scientific-skills","markdown-mermaid-writing",{"evaluate":667,"extract":672},{"promptVersionExtension":205,"promptVersionScoring":206,"score":550,"tags":668,"targetMarket":277,"tier":217},[610,214,669,670,671],"mermaid","scientific-writing","diagramming",{"commitSha":279,"license":673},"Apache-2.0",{"repoId":675},"kd79rphh5gexy91xmpxc05h5mh86mm9r",[671,610,214,669,670],{"evaluatedAt":678,"extractAt":679,"updatedAt":678},1778693085426,1778691799740]