[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"extension-plugin-juliusbrussee-caveman-en":3,"guides-for-juliusbrussee-caveman":622,"similar-k170g496f5ze5hv8xa2e7b91ex86mz9m-en":623},{"_creationTime":4,"_id":5,"children":6,"community":163,"display":164,"evaluation":168,"identity":408,"isFallback":391,"parentExtension":411,"providers":437,"relations":443,"repo":444,"tags":619,"workflow":620},1778691765524.373,"k170g496f5ze5hv8xa2e7b91ex86mz9m",[7,46,69,94,118,142],{"_creationTime":8,"_id":9,"community":10,"display":12,"identity":18,"providers":23,"relations":39,"tags":41,"workflow":42},1778691765524.3735,"k175qp58x34qgtcj9b1tpxeern86mqvz",{"reviewCount":11},0,{"description":13,"installMethods":14,"name":16,"sourceUrl":17},"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":15},"juliusbrussee/caveman","caveman","https://github.com/juliusbrussee/caveman",{"basePath":19,"githubOwner":20,"githubRepo":16,"locale":21,"slug":16,"type":22},"skills/caveman","juliusbrussee","en","skill",{"evaluate":24,"extract":36},{"promptVersionExtension":25,"promptVersionScoring":26,"score":27,"tags":28,"targetMarket":34,"tier":35},"3.0.0","4.4.0",99,[29,30,31,32,33],"compression","llm","token-efficiency","communication","prompt-engineering","global","verified",{"commitSha":37,"license":38},"HEAD","MIT",{"parentExtensionId":5,"repoId":40},"kd7890f0sfy0ymf1pakpx7vqch86nmew",[32,29,30,33,31],{"evaluatedAt":43,"extractAt":44,"updatedAt":45},1778691864047,1778691765524,1778692174530,{"_creationTime":47,"_id":48,"community":49,"display":50,"identity":54,"providers":56,"relations":64,"tags":65,"workflow":66},1778691765524.3738,"k176a757cfd35jb9npje9qqzed86nvpn",{"reviewCount":11},{"description":51,"installMethods":52,"name":53,"sourceUrl":17},"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":15},"caveman-compress",{"basePath":55,"githubOwner":20,"githubRepo":16,"locale":21,"slug":53,"type":22},"skills/caveman-compress",{"evaluate":57,"extract":63},{"promptVersionExtension":25,"promptVersionScoring":26,"score":27,"tags":58,"targetMarket":34,"tier":35},[29,59,60,61,62],"token-saving","markdown","text-processing","utility",{"commitSha":37},{"parentExtensionId":5,"repoId":40},[29,60,61,59,62],{"evaluatedAt":67,"extractAt":44,"updatedAt":68},1778691883875,1778692174897,{"_creationTime":70,"_id":71,"community":72,"display":73,"identity":77,"providers":79,"relations":89,"tags":90,"workflow":91},1778691765524.374,"k1713vd42cpj5hpr0a1vvr8rg986nzre",{"reviewCount":11},{"description":74,"installMethods":75,"name":76,"sourceUrl":17},"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",{"claudeCode":15},"caveman-stats",{"basePath":78,"githubOwner":20,"githubRepo":16,"locale":21,"slug":76,"type":22},"skills/caveman-stats",{"evaluate":80,"extract":88},{"promptVersionExtension":25,"promptVersionScoring":26,"score":81,"tags":82,"targetMarket":34,"tier":87},96,[83,84,85,86],"developer-tools","code-analysis","logging","monitoring","community",{"commitSha":37},{"parentExtensionId":5,"repoId":40},[84,83,85,86],{"evaluatedAt":92,"extractAt":44,"updatedAt":93},1778691906284,1778692174731,{"_creationTime":95,"_id":96,"community":97,"display":98,"identity":102,"providers":104,"relations":113,"tags":114,"workflow":115},1778691765524.3743,"k17bwzaz0dtqsqrjc1gqvqq22s86mcth",{"reviewCount":11},{"description":99,"installMethods":100,"name":101,"sourceUrl":17},"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",{"claudeCode":15},"caveman-commit",{"basePath":103,"githubOwner":20,"githubRepo":16,"locale":21,"slug":101,"type":22},"skills/caveman-commit",{"evaluate":105,"extract":112},{"promptVersionExtension":25,"promptVersionScoring":26,"score":106,"tags":107,"targetMarket":34,"tier":35},98,[108,109,110,83,111],"git","commit","conventional-commits","code-generation",{"commitSha":37},{"parentExtensionId":5,"repoId":40},[111,109,110,83,108],{"evaluatedAt":116,"extractAt":44,"updatedAt":117},1778691948611,1778692175229,{"_creationTime":119,"_id":120,"community":121,"display":122,"identity":126,"providers":128,"relations":137,"tags":138,"workflow":139},1778691765524.3745,"k17f0mmqg0eejq9esy4t7hztmn86ntpe",{"reviewCount":11},{"description":123,"installMethods":124,"name":125,"sourceUrl":17},"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",{"claudeCode":15},"caveman-help",{"basePath":127,"githubOwner":20,"githubRepo":16,"locale":21,"slug":125,"type":22},"skills/caveman-help",{"evaluate":129,"extract":136},{"promptVersionExtension":25,"promptVersionScoring":26,"score":130,"tags":131,"targetMarket":34,"tier":35},100,[132,133,134,135,62],"help","reference","documentation","cheat-sheet",{"commitSha":37},{"parentExtensionId":5,"repoId":40},[135,134,132,133,62],{"evaluatedAt":140,"extractAt":44,"updatedAt":141},1778691968065,1778692175051,{"_creationTime":143,"_id":144,"community":145,"display":146,"identity":150,"providers":152,"relations":158,"tags":159,"workflow":160},1778691765524.3748,"k179sc1ycgeh4jwz2ecyvq2d8986naxj",{"reviewCount":11},{"description":147,"installMethods":148,"name":149,"sourceUrl":17},"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",{"claudeCode":15},"caveman-review",{"basePath":151,"githubOwner":20,"githubRepo":16,"locale":21,"slug":149,"type":22},"skills/caveman-review",{"evaluate":153,"extract":157},{"promptVersionExtension":25,"promptVersionScoring":26,"score":106,"tags":154,"targetMarket":34,"tier":35},[155,83,156,33],"code-review","productivity",{"commitSha":37},{"parentExtensionId":5,"repoId":40},[155,83,156,33],{"evaluatedAt":161,"extractAt":44,"updatedAt":162},1778691992008,1778692175404,{"reviewCount":11},{"description":165,"installMethods":166,"name":167,"sourceUrl":17},"Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.",{"claudeCode":16},"Caveman",{"_creationTime":169,"_id":170,"extensionId":5,"locale":21,"result":171,"trustSignals":389,"workflow":406},1778691819811.295,"kn7a2h94zpwtadsame8yyvkmfn86nvd7",{"checks":172,"evaluatedAt":357,"extensionSummary":358,"features":359,"nonGoals":365,"practices":370,"prerequisites":374,"promptVersionExtension":25,"promptVersionScoring":26,"purpose":378,"rationale":379,"score":27,"summary":380,"tags":381,"targetMarket":34,"tier":35,"useCases":383},[173,178,181,184,188,191,196,200,203,206,210,214,217,221,224,227,230,233,236,239,243,247,251,255,259,262,265,268,272,275,278,281,284,287,290,294,298,302,305,309,312,315,318,321,324,327,330,333,336,339,343,346,349,353],{"category":174,"check":175,"severity":176,"summary":177},"Practical Utility","Problem relevance","pass","The description clearly states the problem of high token usage during communication and offers a solution through ultra-compressed speech.",{"category":174,"check":179,"severity":176,"summary":180},"Unique selling proposition","The extension offers a unique approach to LLM communication by drastically reducing token count while preserving technical accuracy, going beyond simple prompt engineering.",{"category":174,"check":182,"severity":176,"summary":183},"Production readiness","The plugin provides a complete lifecycle for its stated use case, offering various modes, installation methods, and integration with different agents and workflows.",{"category":185,"check":186,"severity":176,"summary":187},"Scope","Single responsibility principle","The plugin focuses on text compression for LLM communication and related utilities, maintaining a coherent domain.",{"category":185,"check":189,"severity":176,"summary":190},"Description quality","The displayed description is concise, accurate, and effectively communicates the core value proposition of the extension.",{"category":192,"check":193,"severity":194,"summary":195},"Invocation","Scoped tools","not_applicable","This is a plugin with a single core command and associated skills, not a collection of individual tools with potentially broad scopes.",{"category":197,"check":198,"severity":176,"summary":199},"Documentation","Configuration & parameter reference","Configuration options and usage details, including intensity levels and auto-activation, are well-documented in the README and associated skill files.",{"category":185,"check":201,"severity":176,"summary":202},"Tool naming","Tool and command names like `/caveman`, `/caveman-commit`, and `caveman-compress` are descriptive and relevant to their function.",{"category":185,"check":204,"severity":176,"summary":205},"Minimal I/O surface","The commands and skills primarily operate on file paths or specific text inputs, with clear and focused output formats.",{"category":207,"check":208,"severity":176,"summary":209},"License","License usability","The extension is distributed under the MIT license, which is permissive and widely usable.",{"category":211,"check":212,"severity":176,"summary":213},"Maintenance","Commit recency","The last commit was on May 12, 2026, indicating recent maintenance activity.",{"category":211,"check":215,"severity":194,"summary":216},"Dependency Management","The installer script appears to be self-contained with no external npm dependencies beyond Node.js itself.",{"category":218,"check":219,"severity":194,"summary":220},"Security","Secret Management","The extension does not handle or require any secrets.",{"category":218,"check":222,"severity":176,"summary":223},"Injection","The extension processes text and file paths but does not appear to execute untrusted code or instructions from external sources.",{"category":218,"check":225,"severity":176,"summary":226},"Transitive Supply-Chain Grenades","The extension relies on Node.js and potentially npm for the installer, but core functionality appears bundled and does not fetch external code at runtime.",{"category":218,"check":228,"severity":176,"summary":229},"Sandbox Isolation","The installer and core scripts operate within standard user permissions and do not attempt to modify files outside of expected configuration directories.",{"category":218,"check":231,"severity":176,"summary":232},"Sandbox escape primitives","No detached processes or retry loops around denied tool calls were detected in the scripts.",{"category":218,"check":234,"severity":176,"summary":235},"Data Exfiltration","The extension's functionality is focused on local text compression and does not involve sending user data to third parties.",{"category":218,"check":237,"severity":176,"summary":238},"Hidden Text Tricks","The README and code files do not contain hidden text tricks or malicious steering characters.",{"category":240,"check":241,"severity":176,"summary":242},"Hooks","Opaque code execution","The installer script is plain JavaScript, and the hook implementations appear to be readable Node.js code.",{"category":244,"check":245,"severity":176,"summary":246},"Portability","Structural Assumption","The installer and scripts are designed to be cross-platform and do not make rigid assumptions about project structure beyond standard agent configurations.",{"category":248,"check":249,"severity":176,"summary":250},"Trust","Issues Attention","With 66 issues opened and 70 closed in the last 90 days, the closure rate is approximately 51%, indicating good maintainer engagement.",{"category":252,"check":253,"severity":176,"summary":254},"Versioning","Release Management","The extension's `package.json` and SKILL.md frontmatter specify versioning, and the commit history shows recent updates.",{"category":256,"check":257,"severity":194,"summary":258},"Code Execution","Validation","The primary functionality involves text manipulation and file operations; complex schema validation is not a core requirement.",{"category":218,"check":260,"severity":176,"summary":261},"Unguarded Destructive Operations","The extension's core commands and skills are non-destructive. The installer includes an uninstall option.",{"category":256,"check":263,"severity":176,"summary":264},"Error Handling","The installer and provided scripts appear to handle errors gracefully and provide informative messages.",{"category":256,"check":266,"severity":194,"summary":267},"Logging","The extension's functionality does not involve persistent destructive actions or outbound calls that would necessitate a local audit log.",{"category":269,"check":270,"severity":176,"summary":271},"Compliance","GDPR","The extension operates on local files and text compression, and does not process personal data.",{"category":269,"check":273,"severity":176,"summary":274},"Target market","The extension's functionality is universally applicable and does not have regional or jurisdictional limitations.",{"category":244,"check":276,"severity":176,"summary":277},"Runtime stability","The installer script is written in Node.js, ensuring cross-platform compatibility. Core logic relies on Node.js and standard shell commands.",{"category":197,"check":279,"severity":176,"summary":280},"README","A comprehensive README exists and clearly articulates the extension's purpose, features, and usage.",{"category":185,"check":282,"severity":176,"summary":283},"Tool surface size","The plugin exposes a reasonable number of commands and skills (approx. 6-7 main ones), avoiding bloat.",{"category":192,"check":285,"severity":176,"summary":286},"Overlapping near-synonym tools","The commands and skills have distinct names and purposes, avoiding ambiguity.",{"category":197,"check":288,"severity":176,"summary":289},"Phantom features","All features mentioned in the README and documentation appear to have corresponding implementations or commands.",{"category":291,"check":292,"severity":176,"summary":293},"Install","Installation instruction","Clear installation instructions are provided for various platforms, including copy-pasteable commands and agent-specific guidance.",{"category":295,"check":296,"severity":176,"summary":297},"Errors","Actionable error messages","Error messages in the installer and documentation provide context and suggest remediation steps.",{"category":299,"check":300,"severity":194,"summary":301},"Execution","Pinned dependencies","The installer script is self-contained and does not rely on third-party dependencies that require pinning.",{"category":185,"check":303,"severity":194,"summary":304},"Dry-run preview","The extension's primary functions are text compression and message generation, which are not state-changing or destructive operations requiring a dry-run.",{"category":306,"check":307,"severity":194,"summary":308},"Protocol","Idempotent retry & timeouts","The core operations are local and do not involve remote calls or state-changing operations requiring idempotency or timeouts.",{"category":269,"check":310,"severity":176,"summary":311},"Telemetry opt-in","The extension does not appear to emit any telemetry by default; any potential telemetry would be opt-in and documented.",{"category":192,"check":313,"severity":176,"summary":314},"Name collisions","The bundled commands and skills have distinct names (e.g., `/caveman`, `/caveman-commit`, `caveman-compress`) and do not appear to collide with built-in Claude Code commands.",{"category":192,"check":316,"severity":176,"summary":317},"Hooks-off mechanism","The extension provides clear instructions on how to disable caveman mode using commands like 'stop caveman' or 'normal mode', and documentation on manual deactivation.",{"category":192,"check":319,"severity":194,"summary":320},"Hook matcher tightness","The hooks appear to be triggered by specific commands or events (SessionStart, UserPromptSubmit) rather than broad tool matching.",{"category":218,"check":322,"severity":176,"summary":323},"Hook security","The hooks are primarily for mode management and statusline updates, not destructive or network-touching operations.",{"category":240,"check":325,"severity":194,"summary":326},"Silent prompt rewriting","The `UserPromptSubmit` hook appears to be for tracking mode rather than rewriting prompts, and its actions are likely transparent to the user.",{"category":218,"check":328,"severity":194,"summary":329},"Permission Hook","There are no `PermissionRequest` hooks present in the plugin configuration.",{"category":269,"check":331,"severity":176,"summary":332},"Hook privacy","The hooks are designed for local mode management and statusline updates, without sending data over the network.",{"category":256,"check":334,"severity":176,"summary":335},"Hook dependency","The hooks are implemented as short, readable Node.js scripts within the repository.",{"category":197,"check":337,"severity":176,"summary":338},"Feature Transparency","Critical functionality, including hooks and commands, is described in the README and associated SKILL.md files.",{"category":340,"check":341,"severity":176,"summary":342},"Convention","Layout convention adherence","The plugin follows standard Claude Code plugin structure, with hooks managed appropriately and related scripts in `bin/`.",{"category":340,"check":344,"severity":176,"summary":345},"Plugin state","Plugin state, if any, is managed through configuration files and local scripts, adhering to standard conventions.",{"category":218,"check":347,"severity":194,"summary":348},"Keychain-stored secrets","The extension does not handle secrets that would require keychain storage.",{"category":350,"check":351,"severity":176,"summary":352},"Dependencies","Tagged release sourcing","The installer script references the main branch of the GitHub repository, which is actively maintained.",{"category":354,"check":355,"severity":176,"summary":356},"Installation","Clean uninstall","The uninstaller script appears to cleanly remove installed components without leaving background processes or persistent artifacts.",1778691819702,"This plugin drastically reduces token usage in LLM communication by making the agent speak like a caveman, while preserving technical accuracy. It offers various intensity levels and includes commands for commit messages, code reviews, file compression, and stats.",[360,361,362,363,364],"Ultra-compressed communication mode","Multiple intensity levels (lite, full, ultra, wenyan)","Dedicated skills for commit messages, code reviews, and file compression","Automatic statusline integration and session activation","Cross-platform installer",[366,367,368,369],"Reducing the LLM's internal thinking/reasoning tokens","Altering the LLM's core knowledge or capabilities","Providing complex code editing or refactoring features beyond simple edits","Replacing the LLM's default verbose output for sensitive or complex explanations",[371,372,373],"Code compression","LLM interaction optimization","Developer productivity",[375,376,377],"Node.js ≥18","Claude Code or compatible agent","Optional: `git` for commit/review features","To significantly reduce LLM token consumption and improve response speed by employing a compressed, caveman-like communication style.","The extension has a clear value proposition, excellent documentation, and recent maintenance. All checks passed or were not applicable, indicating high quality.","An effective plugin for compressing LLM communication, saving tokens and improving response speed.",[29,30,382,83,156],"text",[384,385,386,387,388],"Reducing LLM API costs by minimizing token usage","Speeding up LLM response times","Generating concise commit messages and code review comments","Compressing local memory files to fit more context","Integrating with various LLM agents and IDEs",{"codeQuality":390,"collectedAt":392,"documentation":393,"maintenance":396,"popularity":402,"security":403,"testCoverage":405},{"hasLockfile":391},false,1778691801610,{"descriptionLength":394,"readmeSize":395},122,10896,{"closedIssues90d":397,"forks":398,"hasChangelog":391,"openIssues90d":399,"pushedAt":400,"stars":401},70,3292,66,1778615088000,59509,{"npmDownloads":11},{"hasNpmPackage":404,"license":38,"smitheryVerified":391},true,{"hasCi":404,"hasTests":404},{"updatedAt":407},1778691819811,{"basePath":409,"githubOwner":20,"githubRepo":16,"locale":21,"slug":16,"type":410},"","plugin",{"_creationTime":412,"_id":413,"community":414,"display":415,"identity":418,"parentExtension":420,"providers":421,"relations":432,"tags":433,"workflow":434},1778691765524.3728,"k17dcfwpwpzjc43zt2qawvft3h86mmse",{"reviewCount":11},{"description":416,"installMethods":417,"name":16,"sourceUrl":17},"Ultra-compressed communication mode for Claude Code. Cuts ~75% of tokens while keeping full technical accuracy.",{"claudeCode":15},{"basePath":409,"githubOwner":20,"githubRepo":16,"locale":21,"slug":16,"type":419},"marketplace",null,{"evaluate":422,"extract":427},{"promptVersionExtension":423,"promptVersionScoring":26,"score":424,"tags":425,"targetMarket":34,"tier":35},"3.1.0",92,[29,30,156,426,83],"token-reduction",{"commitSha":37,"marketplace":428,"plugin":430},{"name":16,"pluginCount":429},1,{"mcpCount":11,"provider":431,"skillCount":11},"classify",{"repoId":40},[29,83,30,156,426],{"evaluatedAt":435,"extractAt":44,"updatedAt":436},1778691801317,1778692173865,{"evaluate":438,"extract":440},{"promptVersionExtension":25,"promptVersionScoring":26,"score":27,"tags":439,"targetMarket":34,"tier":35},[29,30,382,83,156],{"commitSha":37,"license":38,"plugin":441},{"mcpCount":11,"provider":431,"skillCount":442},6,{"parentExtensionId":413,"repoId":40},{"_creationTime":445,"_id":40,"identity":446,"providers":447,"workflow":614},1778691760140.7458,{"githubOwner":20,"githubRepo":16,"sourceUrl":17},{"classify":448,"discover":599,"extract":602,"github":603,"npm":613},{"commitSha":37,"extensions":449},[450,460,471,515,524,530,552,558,564,570,576,588],{"basePath":451,"displayName":452,"installMethods":453,"rationale":454,"selectedPaths":455,"source":459,"sourceLanguage":21,"type":419},".agents/plugins","caveman-repo",{"claudeCode":15},"marketplace.json at .agents/plugins/marketplace.json",[456],{"path":457,"priority":458},"marketplace.json","mandatory","rule",{"basePath":409,"description":416,"displayName":16,"installMethods":461,"rationale":462,"selectedPaths":463,"source":459,"sourceLanguage":21,"type":419},{"claudeCode":15},"marketplace.json at .claude-plugin/marketplace.json",[464,466,468],{"path":465,"priority":458},".claude-plugin/marketplace.json",{"path":467,"priority":458},"README.md",{"path":469,"priority":470},"LICENSE","high",{"basePath":409,"description":165,"displayName":16,"installMethods":472,"rationale":473,"selectedPaths":474,"source":459,"sourceLanguage":21,"type":410},{"claudeCode":16},"plugin manifest at .claude-plugin/plugin.json (coalesced with duplicate plugin at plugins/caveman)",[475,477,478,479,482,484,486,488,490,492,494,496,498,500,503,505,507,509,511,513],{"path":476,"priority":458},".claude-plugin/plugin.json",{"path":467,"priority":458},{"path":469,"priority":470},{"path":480,"priority":481},"skills/cavecrew/SKILL.md","medium",{"path":483,"priority":481},"skills/caveman/SKILL.md",{"path":485,"priority":481},"skills/caveman-commit/SKILL.md",{"path":487,"priority":481},"skills/caveman-compress/SKILL.md",{"path":489,"priority":481},"skills/caveman-help/SKILL.md",{"path":491,"priority":481},"skills/caveman-review/SKILL.md",{"path":493,"priority":481},"skills/caveman-stats/SKILL.md",{"path":495,"priority":470},"agents/cavecrew-builder.md",{"path":497,"priority":470},"agents/cavecrew-investigator.md",{"path":499,"priority":470},"agents/cavecrew-reviewer.md",{"path":501,"priority":502},"bin/install.js","low",{"path":504,"priority":502},"bin/lib/openclaw.js",{"path":506,"priority":502},"bin/lib/settings.js",{"path":508,"priority":470},"commands/caveman-commit.toml",{"path":510,"priority":470},"commands/caveman-init.toml",{"path":512,"priority":470},"commands/caveman-review.toml",{"path":514,"priority":470},"commands/caveman.toml",{"basePath":516,"description":517,"displayName":518,"installMethods":519,"rationale":520,"selectedPaths":521,"source":459,"sourceLanguage":21,"type":22},".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":15},"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)",[522],{"path":523,"priority":458},"SKILL.md",{"basePath":19,"description":13,"displayName":16,"installMethods":525,"rationale":526,"selectedPaths":527,"source":459,"sourceLanguage":21,"type":22},{"claudeCode":15},"SKILL.md frontmatter at skills/caveman/SKILL.md (coalesced with duplicate skill at plugins/caveman/skills/caveman)",[528,529],{"path":523,"priority":458},{"path":467,"priority":470},{"basePath":55,"description":51,"displayName":53,"installMethods":531,"rationale":532,"selectedPaths":533,"source":459,"sourceLanguage":21,"type":22},{"claudeCode":15},"SKILL.md frontmatter at skills/caveman-compress/SKILL.md (coalesced with duplicate skill at plugins/caveman/skills/caveman-compress)",[534,535,536,538,540,542,544,546,548,550],{"path":523,"priority":458},{"path":467,"priority":470},{"path":537,"priority":481},"SECURITY.md",{"path":539,"priority":502},"scripts/__init__.py",{"path":541,"priority":502},"scripts/__main__.py",{"path":543,"priority":502},"scripts/benchmark.py",{"path":545,"priority":502},"scripts/cli.py",{"path":547,"priority":502},"scripts/compress.py",{"path":549,"priority":502},"scripts/detect.py",{"path":551,"priority":502},"scripts/validate.py",{"basePath":78,"description":74,"displayName":76,"installMethods":553,"rationale":554,"selectedPaths":555,"source":459,"sourceLanguage":21,"type":22},{"claudeCode":15},"SKILL.md frontmatter at skills/caveman-stats/SKILL.md (coalesced with duplicate skill at plugins/caveman/skills/caveman-stats)",[556,557],{"path":523,"priority":458},{"path":467,"priority":470},{"basePath":103,"description":99,"displayName":101,"installMethods":559,"rationale":560,"selectedPaths":561,"source":459,"sourceLanguage":21,"type":22},{"claudeCode":15},"SKILL.md frontmatter at skills/caveman-commit/SKILL.md",[562,563],{"path":523,"priority":458},{"path":467,"priority":470},{"basePath":127,"description":123,"displayName":125,"installMethods":565,"rationale":566,"selectedPaths":567,"source":459,"sourceLanguage":21,"type":22},{"claudeCode":15},"SKILL.md frontmatter at skills/caveman-help/SKILL.md",[568,569],{"path":523,"priority":458},{"path":467,"priority":470},{"basePath":151,"description":147,"displayName":149,"installMethods":571,"rationale":572,"selectedPaths":573,"source":459,"sourceLanguage":21,"type":22},{"claudeCode":15},"SKILL.md frontmatter at skills/caveman-review/SKILL.md",[574,575],{"path":523,"priority":458},{"path":467,"priority":470},{"basePath":409,"description":577,"displayName":578,"installMethods":579,"license":38,"rationale":580,"selectedPaths":581,"source":459,"sourceLanguage":21,"type":587},"Caveman installer — detects your AI coding agents and installs caveman for each one.","caveman-installer",{"npm":578},"cli ecosystem detected at /",[582,584,585,586],{"path":583,"priority":458},"package.json",{"path":467,"priority":458},{"path":469,"priority":470},{"path":501,"priority":481},"cli",{"basePath":589,"description":590,"displayName":591,"installMethods":592,"license":38,"rationale":593,"selectedPaths":594,"source":459,"sourceLanguage":21,"type":587},"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":591},"cli ecosystem detected at src/mcp-servers/caveman-shrink",[595,596,597],{"path":583,"priority":458},{"path":467,"priority":458},{"path":598,"priority":481},"index.js",{"sources":600},[601],"manual",{"npmPackage":578},{"closedIssues90d":397,"description":604,"forks":398,"homepage":605,"license":38,"openIssues90d":399,"pushedAt":400,"readmeSize":395,"stars":401,"topics":606},"🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman","https://getcaveman.dev/",[607,608,16,609,610,30,611,33,22,612],"ai","anthropic","claude","claude-code","meme","tokens",{"downloads":11},{"classifiedAt":615,"discoverAt":616,"extractAt":617,"githubAt":617,"npmAt":618,"updatedAt":615},1778691765329,1778691760140,1778691762475,1778691764072,[29,83,30,156,382],{"evaluatedAt":407,"extractAt":44,"updatedAt":621},1778692174193,[],[624,651,677,705,735,762],{"_creationTime":625,"_id":626,"community":627,"display":628,"identity":633,"providers":637,"relations":643,"tags":646,"workflow":647},1778698235845.4077,"k17f8b1e3611rh6d9e6peh43b186m55k",{"reviewCount":11},{"description":629,"installMethods":630,"name":631,"sourceUrl":632},"Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.",{"claudeCode":631},"context7-plugin","https://github.com/upstash/context7",{"basePath":634,"githubOwner":635,"githubRepo":636,"locale":21,"slug":636,"type":410},"plugins/claude/context7","upstash","context7",{"evaluate":638,"extract":642},{"promptVersionExtension":25,"promptVersionScoring":26,"score":130,"tags":639,"targetMarket":34,"tier":35},[134,83,640,641],"code-examples","mcp-server",{"commitSha":37},{"parentExtensionId":644,"repoId":645},"k17c6qmv4dnjycsp8aa4wyfbgh86n3jd","kd7955sg5wbf89gw527wdep66n86na9w",[640,83,134,641],{"evaluatedAt":648,"extractAt":649,"updatedAt":650},1778698268645,1778698235845,1778698495329,{"_creationTime":652,"_id":653,"community":654,"display":655,"identity":660,"providers":663,"relations":670,"tags":672,"workflow":673},1778692816134.1846,"k17433617dh29ajy4hx5q8txzs86nvg4",{"reviewCount":11},{"description":656,"installMethods":657,"name":658,"sourceUrl":659},"Markdown-based Kanban board managed by Claude Code. Cards live as .md files — no database, no server.",{"claudeCode":658},"kanban","https://github.com/mattjoyce/kanban-skill",{"basePath":409,"githubOwner":661,"githubRepo":662,"locale":21,"slug":662,"type":410},"mattjoyce","kanban-skill",{"evaluate":664,"extract":667},{"promptVersionExtension":25,"promptVersionScoring":26,"score":130,"tags":665,"targetMarket":34,"tier":35},[658,60,666,156,587],"task-management",{"commitSha":37,"license":668,"plugin":669},"Apache-2.0",{"mcpCount":11,"provider":431,"skillCount":429},{"repoId":671},"kd704ngckmedtqtqt31s01j8kx86mryd",[587,658,60,156,666],{"evaluatedAt":674,"extractAt":675,"updatedAt":676},1778692838399,1778692816134,1778692888073,{"_creationTime":678,"_id":679,"community":680,"display":681,"identity":686,"providers":688,"relations":698,"tags":700,"workflow":701},1778692275423.988,"k17ccytjrakavnrzgdbzkbcjj186nr1f",{"reviewCount":11},{"description":682,"installMethods":683,"name":684,"sourceUrl":685},"Makes Claude actually follow your rules. Save rules with natural language, enforce them with hooks, remember them across sessions.",{"claudeCode":684},"obey","https://github.com/Lexxes-Projects/obey",{"basePath":409,"githubOwner":687,"githubRepo":684,"locale":21,"slug":684,"type":410},"Lexxes-Projects",{"evaluate":689,"extract":695},{"promptVersionExtension":25,"promptVersionScoring":26,"score":130,"tags":690,"targetMarket":34,"tier":35},[691,692,693,694,156],"rules","enforcement","llm-behavior","ai-assistant",{"commitSha":37,"license":38,"plugin":696},{"mcpCount":11,"provider":431,"skillCount":697},3,{"repoId":699},"kd75x7p4gkt5vx785rmc3msdhs86neap",[694,692,693,156,691],{"evaluatedAt":702,"extractAt":703,"updatedAt":704},1778692293423,1778692275424,1778692422045,{"_creationTime":706,"_id":707,"community":708,"display":709,"identity":714,"providers":716,"relations":727,"tags":730,"workflow":731},1778693661691.4358,"k177fsagh49r77m9y4755zc1mn86m1jm",{"reviewCount":11},{"description":710,"installMethods":711,"name":712,"sourceUrl":713},"Make assistant output sound human. Strip AI-isms (sycophancy, stock vocab, hedging stacks, em-dash pileups), engineer burstiness, restore voice. Preserves code, URLs, and technical accuracy.",{"claudeCode":712},"unslop","https://github.com/MohamedAbdallah-14/unslop",{"basePath":409,"githubOwner":715,"githubRepo":712,"locale":21,"slug":712,"type":410},"MohamedAbdallah-14",{"evaluate":717,"extract":724},{"promptVersionExtension":25,"promptVersionScoring":26,"score":130,"tags":718,"targetMarket":34,"tier":35},[607,382,719,720,721,722,723],"writing","editor","code","nlp","humanizer",{"commitSha":37,"plugin":725},{"mcpCount":11,"provider":431,"skillCount":726},5,{"parentExtensionId":728,"repoId":729},"k175vxsqnmn2ye2xkw62x4enkh86n8eb","kd727xcarpnqcat3wd68ms466s86mwkb",[607,721,720,723,722,382,719],{"evaluatedAt":732,"extractAt":733,"updatedAt":734},1778693722676,1778693661691,1778693923675,{"_creationTime":736,"_id":737,"community":738,"display":739,"identity":744,"providers":748,"relations":755,"tags":758,"workflow":759},1778696052276.008,"k173c9f6h756g95dhcd0eb0g6n86n2jw",{"reviewCount":11},{"description":740,"installMethods":741,"name":742,"sourceUrl":743},"Claude Code expertise: skills, commands, hooks, MCP, settings (7 skills)",{"claudeCode":608},"ai-skills","https://github.com/raintree-technology/claude-starter",{"basePath":745,"githubOwner":746,"githubRepo":747,"locale":21,"slug":608,"type":410},"templates/.claude/skills/anthropic","raintree-technology","claude-starter",{"evaluate":749,"extract":754},{"promptVersionExtension":25,"promptVersionScoring":26,"score":27,"tags":750,"targetMarket":34,"tier":35},[751,608,83,587,752,29,753],"fintech","json","api",{"commitSha":37},{"parentExtensionId":756,"repoId":757},"k17ab1tq0pfjcwj6czghvnwvzn86mye7","kd78ywakatnz4sjfx781sy14vh86mtty",[608,753,587,29,83,751,752],{"evaluatedAt":760,"extractAt":761,"updatedAt":760},1778696086128,1778696052276,{"_creationTime":763,"_id":764,"community":765,"display":766,"identity":770,"providers":772,"relations":779,"tags":780,"workflow":781},1778696052276.0093,"k175htk3qfrpb932e7vegf4ajx86m22f",{"reviewCount":11},{"description":767,"installMethods":768,"name":769,"sourceUrl":743},"TOON format for 30-60% token savings on tabular data",{"claudeCode":769},"toon-formatter",{"basePath":771,"githubOwner":746,"githubRepo":747,"locale":21,"slug":769,"type":410},"templates/.claude/skills/toon-formatter",{"evaluate":773,"extract":778},{"promptVersionExtension":25,"promptVersionScoring":26,"score":774,"tags":775,"targetMarket":34,"tier":35},97,[776,752,29,83,587,777],"data-formatting","performance",{"commitSha":37},{"parentExtensionId":756,"repoId":757},[587,29,776,83,752,777],{"evaluatedAt":782,"extractAt":761,"updatedAt":782},1778696214208]