{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"github-juliusbrussee--caveman","slug":"juliusbrussee--caveman","name":"caveman","type":"skill","url":"https://getcaveman.dev/","page_url":"https://unfragile.ai/juliusbrussee--caveman","categories":["chatbots-assistants"],"tags":["ai","anthropic","caveman","claude","claude-code","llm","meme","prompt-engineering","skill","tokens"],"pricing":{"model":"open_source","free":true,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"github-juliusbrussee--caveman__cap_0","uri":"capability://text.generation.language.linguistic.token.compression.via.rule.based.transformation","name":"linguistic-token-compression-via-rule-based-transformation","description":"Applies a multi-intensity rule engine (Lite/Full/Ultra modes) that surgically removes linguistic filler—articles, hedging phrases, pleasantries—while preserving code blocks, technical terminology, and safety-critical information. Uses a single-source-of-truth SKILL.md configuration file that defines transformation rules across all host environments (Claude Code, Codex, Gemini CLI), achieving ~75% token reduction without sacrificing technical accuracy through a 'Smart Caveman' principle that protects machine-critical data.","intents":["Reduce API costs by cutting token consumption on repetitive Claude interactions","Accelerate response generation speed by constraining model verbosity","Remove 'walls of text' while maintaining 100% technical clarity in code explanations","Enforce consistent ultra-compressed communication style across multiple AI agent platforms"],"best_for":["Teams running high-volume Claude Code sessions with cost-sensitive budgets","Developers building LLM agents that need deterministic, token-efficient outputs","Organizations migrating from verbose to ultra-compressed AI communication patterns"],"limitations":["Intensity levels (Lite/Full/Ultra) are globally scoped—cannot apply different compression ratios to different conversation sections without manual toggling","Rule engine is linguistic-pattern-based, not semantic—may over-compress domain-specific jargon if not explicitly whitelisted in SKILL.md","No built-in context awareness for when compression should be disabled (e.g., user-facing documentation generation)","Requires explicit activation per session; no persistent state across Claude Code instances"],"requires":["Claude Code plugin infrastructure OR Codex CLI OR Gemini CLI","Access to SKILL.md configuration file (editable by users)","Node.js 14+ for npx skill installation method","Python 3.7+ for benchmarking/testing utilities"],"input_types":["natural language prompts","code snippets (preserved as-is)","markdown-formatted text","multi-turn conversation context"],"output_types":["compressed natural language text","code blocks (unchanged)","structured markdown with reduced prose","technical explanations in caveman-speak"],"categories":["text-generation-language","prompt-engineering"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-juliusbrussee--caveman__cap_1","uri":"capability://tool.use.integration.multi.platform.skill.plugin.deployment","name":"multi-platform-skill-plugin-deployment","description":"Distributes caveman as a portable 'Skill' artifact across heterogeneous AI agent platforms (Claude Code plugin marketplace, Codex CLI, Gemini CLI) using a unified SKILL.md configuration format. Provides platform-specific installation hooks (shell scripts for macOS/Linux/WSL, PowerShell for Windows) that auto-merge configuration into host environment settings (~/.claude/settings.json, Codex config, etc.), enabling single-source-of-truth behavior across all platforms without duplicating rule definitions.","intents":["Deploy caveman across multiple AI agent platforms without maintaining separate codebases","Automatically configure caveman on new machines via shell/PowerShell hooks","Synchronize caveman behavior rules across Claude Code, Codex, and Gemini CLI instances","Enable users to install caveman via npx, plugin marketplace, or manual CLI setup"],"best_for":["Teams using multiple AI agent platforms (Claude Code + Codex + Gemini) simultaneously","DevOps/automation engineers setting up caveman across fleet of developer machines","Open-source maintainers distributing skills across fragmented plugin ecosystems"],"limitations":["SKILL.md sync workflow requires manual GitHub Actions trigger or CI/CD integration—no real-time propagation to installed instances","Platform-specific configuration paths differ (~/.claude/settings.json vs Codex config vs Gemini CLI)—installation scripts must be maintained per OS/platform combination","No rollback mechanism if SKILL.md update breaks existing workflows; users must manually revert configuration","Plugin marketplace approval process (Claude Code) adds latency to distribution; Codex/Gemini CLI updates may lag behind GitHub releases"],"requires":["Node.js 14+ for npx skill installation","Bash/Zsh (macOS/Linux/WSL) OR PowerShell 5.0+ (Windows)","Write access to ~/.claude/settings.json or equivalent platform config directory","Git (for cloning repository and CI/CD workflows)","GitHub Actions (for SKILL.md sync and GitHub Pages deployment)"],"input_types":["SKILL.md configuration file","platform-specific installation scripts (.sh, .ps1)","plugin.json manifest (Claude Code)","GitHub repository metadata"],"output_types":["installed skill artifact in platform-specific location","merged configuration in host environment settings","GitHub Pages documentation site","CI/CD deployment logs"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-juliusbrussee--caveman__cap_2","uri":"capability://text.generation.language.intensity.level.based.compression.tuning","name":"intensity-level-based-compression-tuning","description":"Exposes three discrete compression intensity levels (Lite, Full, Ultra) that users can toggle per session, each applying progressively aggressive linguistic transformation rules. Lite mode removes only obvious filler (articles, some hedging); Full mode aggressively compresses prose while preserving code and technical terms; Ultra mode maximizes compression by removing even more linguistic scaffolding. Implementation uses a rule registry in SKILL.md that maps intensity levels to specific transformation patterns, allowing users to trade off readability vs. token savings without code changes.","intents":["Allow users to dynamically adjust compression aggressiveness based on task complexity (e.g., Lite for exploratory prompts, Ultra for cost-sensitive batch operations)","Provide fallback to lower intensity if Ultra mode over-compresses and loses clarity","Enable A/B testing of compression levels to find optimal token/clarity tradeoff for specific workflows","Support gradual adoption of caveman-speak by starting with Lite and escalating to Full/Ultra"],"best_for":["Developers iterating on prompt engineering who need fine-grained control over compression","Teams running mixed workloads (some requiring clarity, others requiring cost optimization)","Researchers benchmarking token savings vs. output quality across intensity levels"],"limitations":["Intensity levels are binary toggles per session—no mid-conversation switching without manual command invocation","No automatic intensity selection based on task type or context length; users must manually choose level","Ultra mode may produce output that is technically accurate but difficult for humans to parse without training","No metrics/feedback loop to suggest optimal intensity for a given prompt; users must experiment manually"],"requires":["Claude Code, Codex, or Gemini CLI with caveman skill installed","Knowledge of intensity level semantics (Lite/Full/Ultra) and their tradeoffs","Ability to invoke caveman commands to switch intensity mid-session"],"input_types":["intensity level selection (Lite/Full/Ultra)","natural language prompts","code snippets"],"output_types":["compressed text at selected intensity level","code blocks (unchanged)","technical explanations in caveman-speak"],"categories":["text-generation-language","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-juliusbrussee--caveman__cap_3","uri":"capability://safety.moderation.technical.term.and.code.block.preservation","name":"technical-term-and-code-block-preservation","description":"Implements a whitelist-based protection mechanism that exempts code blocks (markdown fences), technical terminology (e.g., useMemo, shallow comparison), and safety-critical operations (e.g., rm -rf) from compression rules. Uses pattern matching and AST-aware detection to identify protected regions, ensuring that compression never degrades technical accuracy or introduces ambiguity in destructive commands. This 'Smart Caveman' principle is enforced via explicit rules in SKILL.md that define protected patterns and categories.","intents":["Ensure code examples remain syntactically correct and unmodified during compression","Preserve domain-specific technical jargon that would lose meaning if compressed","Protect destructive operations (rm -rf, DROP TABLE, etc.) from ambiguous compression","Maintain 100% technical accuracy guarantee while achieving 75% token savings"],"best_for":["Technical documentation and code explanation workflows where accuracy is non-negotiable","Safety-critical operations (DevOps, database migrations) where ambiguity could cause harm","Teams requiring certified technical accuracy alongside cost optimization"],"limitations":["Whitelist maintenance burden—new technical terms or patterns must be explicitly added to SKILL.md to be protected","Pattern matching for code blocks relies on markdown fence detection; non-standard code formatting may not be recognized","Safety-critical operation detection is pattern-based (regex), not semantic—may miss novel destructive operations not in whitelist","No automatic learning or feedback loop to suggest new protected patterns based on user corrections"],"requires":["Caveman skill installed with SKILL.md containing protection rules","Markdown-formatted code blocks (triple backticks) for code preservation","Explicit whitelist entries in SKILL.md for domain-specific technical terms"],"input_types":["natural language text with embedded code blocks","technical explanations with domain-specific jargon","destructive command examples"],"output_types":["compressed text with code blocks unchanged","preserved technical terminology","unambiguous destructive operation descriptions"],"categories":["safety-moderation","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-juliusbrussee--caveman__cap_4","uri":"capability://data.processing.analysis.benchmark.and.performance.measurement.framework","name":"benchmark-and-performance-measurement-framework","description":"Provides a Python-based benchmarking suite (benchmarks/run.py) that measures caveman's token savings, compression ratios, generation speed, and technical accuracy across multiple intensity levels and test prompts. Generates quantitative metrics (e.g., ~75% token savings, ~46% input compression, ~3x speed increase) and supports custom benchmark prompts. Results are published to GitHub Pages documentation, enabling transparent performance tracking and user-facing proof of efficiency gains.","intents":["Quantify token savings and cost reduction achieved by caveman across real-world prompts","Measure generation speed improvements (claimed ~3x faster) with reproducible methodology","Validate that technical accuracy remains at 100% across all intensity levels","Enable users to benchmark caveman on their own prompts and workloads"],"best_for":["Teams evaluating caveman before adoption and needing quantitative ROI metrics","Researchers studying token compression techniques and linguistic transformation","Maintainers tracking performance regressions across caveman releases"],"limitations":["Benchmark prompts are curated and may not represent all real-world use cases; results may vary significantly for domain-specific workloads","Benchmarking requires API calls to Claude (or other LLMs), incurring costs and latency; not suitable for continuous CI/CD runs","Technical accuracy measurement is manual or heuristic-based (not automated); subjective assessment of whether compressed output is 'correct'","Benchmark results are point-in-time snapshots; model behavior changes (Claude updates) may invalidate historical comparisons"],"requires":["Python 3.7+","API key for Claude (or other LLM being benchmarked)","Network access to run benchmarks","Git and GitHub Actions for publishing results to GitHub Pages"],"input_types":["benchmark prompt templates","intensity level configurations","LLM API credentials"],"output_types":["token count metrics (input/output/total)","compression ratio percentages","generation speed measurements (tokens/sec)","technical accuracy assessments","HTML/Markdown benchmark reports for GitHub Pages"],"categories":["data-processing-analysis","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-juliusbrussee--caveman__cap_5","uri":"capability://automation.workflow.github.pages.documentation.and.wiki.generation","name":"github-pages-documentation-and-wiki-generation","description":"Automatically generates and publishes comprehensive documentation to GitHub Pages via CI/CD pipeline, including installation guides, intensity level explanations, linguistic rules, trigger/command reference, plugin architecture details, and benchmark results. Documentation is derived from SKILL.md and repository metadata, ensuring single-source-of-truth consistency. Provides both human-readable guides and technical deep-dives for developers integrating caveman into custom workflows.","intents":["Provide centralized, always-up-to-date documentation without manual wiki maintenance","Enable users to understand caveman's linguistic rules and intensity levels before installation","Document plugin architecture and integration points for developers building custom skills","Publish benchmark results and performance metrics transparently"],"best_for":["Open-source projects needing automated, low-maintenance documentation","Teams distributing skills across multiple platforms and needing unified reference","Users evaluating caveman and needing comprehensive technical documentation"],"limitations":["GitHub Pages deployment requires GitHub Actions CI/CD integration; not suitable for private repositories without GitHub Enterprise","Documentation generation is automated but may require manual review/editing for clarity; no built-in quality gates","Search functionality is limited to GitHub's built-in search; no full-text search index for large documentation sets","Documentation updates are tied to repository commits; real-time documentation updates are not supported"],"requires":["GitHub repository with GitHub Pages enabled","GitHub Actions workflow for CI/CD deployment","Markdown-formatted documentation source files","Jekyll or similar static site generator (GitHub Pages default)"],"input_types":["SKILL.md configuration file","README.md and other markdown documentation","benchmark results (JSON/CSV)","repository metadata (topics, description, etc.)"],"output_types":["HTML documentation site on GitHub Pages","searchable wiki pages","installation guides","API reference documentation","benchmark result visualizations"],"categories":["automation-workflow","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-juliusbrussee--caveman__cap_6","uri":"capability://tool.use.integration.session.scoped.activation.and.command.triggering","name":"session-scoped-activation-and-command-triggering","description":"Provides explicit command-based activation mechanism (e.g., /caveman, /caveman lite, /caveman full, /caveman ultra) that users invoke to enable compression for a specific session. Activation is session-scoped (not persistent across Claude Code instances) and can be toggled on/off mid-conversation. Implementation uses Claude Code's command/trigger system to intercept user input and apply caveman rules to model output, without requiring permanent configuration changes.","intents":["Allow users to enable caveman on-demand without permanent configuration changes","Support A/B testing by toggling caveman on/off within same conversation","Provide explicit control over when compression is applied (e.g., disable for user-facing output)","Enable gradual adoption by letting users test caveman before committing to permanent activation"],"best_for":["Users wanting to test caveman without permanent configuration","Workflows requiring selective compression (some prompts compressed, others not)","Teams piloting caveman adoption across organization"],"limitations":["Session-scoped activation means caveman must be re-enabled for each new Claude Code session; no persistent state across sessions","Command-based triggering requires users to remember and type activation commands; no automatic activation based on context","Mid-conversation toggling adds cognitive overhead; users must manually track which sections are compressed vs. uncompressed","No integration with Claude Code's native settings or preferences; activation is ephemeral"],"requires":["Claude Code plugin with caveman skill installed","Knowledge of caveman activation commands (/caveman, /caveman lite, etc.)","Ability to type commands in Claude Code chat interface"],"input_types":["caveman activation commands","natural language prompts (after activation)"],"output_types":["compressed text output (after activation)","status messages confirming activation/deactivation"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-juliusbrussee--caveman__cap_7","uri":"capability://data.processing.analysis.linguistic.rule.registry.and.pattern.matching","name":"linguistic-rule-registry-and-pattern-matching","description":"Implements a declarative rule registry in SKILL.md that defines linguistic transformation patterns (e.g., 'The reason is' → 'Reason:', delete articles 'a'/'an'/'the', remove hedging phrases). Rules are organized by category (grammar, articles, filler, safety-critical) and intensity level (Lite/Full/Ultra), enabling pattern-based text transformation without hardcoded logic. Uses regex or string-matching patterns to identify and replace linguistic elements, with explicit exceptions for code blocks and technical terms.","intents":["Define compression rules in a maintainable, declarative format that non-engineers can understand and modify","Enable community contributions to caveman by lowering barrier to adding new linguistic rules","Support version control and auditing of rule changes via SKILL.md git history","Allow users to customize compression rules for domain-specific terminology or communication styles"],"best_for":["Teams wanting to customize caveman rules for domain-specific jargon","Open-source communities contributing new linguistic patterns","Organizations auditing AI communication styles and needing rule transparency"],"limitations":["Pattern matching is regex/string-based, not semantic—may produce false positives/negatives (e.g., 'a' in 'data' vs. article 'a')","Rule ordering matters; overlapping patterns may conflict; no built-in conflict resolution","No automatic validation of rules; malformed patterns may silently fail or produce unexpected output","Rule complexity grows with intensity levels; Full/Ultra modes may have hundreds of patterns, making maintenance difficult","No rollback mechanism if rule change breaks existing workflows; users must manually revert SKILL.md"],"requires":["Text editor for editing SKILL.md","Understanding of regex syntax (for pattern-based rules)","Knowledge of caveman's rule categories and intensity levels"],"input_types":["SKILL.md configuration file with rule definitions","natural language text to be transformed"],"output_types":["transformed text with rules applied","rule validation reports (if implemented)"],"categories":["data-processing-analysis","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-juliusbrussee--caveman__cap_8","uri":"capability://automation.workflow.cross.platform.configuration.synchronization","name":"cross-platform-configuration-synchronization","description":"Implements a GitHub Actions CI/CD workflow that synchronizes SKILL.md configuration changes across all supported platforms (Claude Code, Codex, Gemini CLI) by automatically regenerating platform-specific configuration files and triggering plugin updates. Uses a single-source-of-truth SKILL.md that is parsed and transformed into platform-native formats (plugin.json for Claude Code, config files for Codex/Gemini), ensuring behavior consistency without manual per-platform updates.","intents":["Maintain consistent caveman behavior across Claude Code, Codex, and Gemini CLI without manual synchronization","Automate platform-specific configuration generation from single SKILL.md source","Enable rapid rule updates that propagate to all platforms via CI/CD pipeline","Reduce maintenance burden by eliminating duplicate configuration across platforms"],"best_for":["Maintainers distributing caveman across multiple AI agent platforms","Teams requiring consistent behavior across heterogeneous tooling","Organizations with strict configuration management requirements"],"limitations":["CI/CD sync workflow requires GitHub Actions; not suitable for self-hosted or private repositories without GitHub Enterprise","Platform-specific configuration formats differ significantly; transformation logic must be maintained for each platform","Sync latency: updates to SKILL.md may take minutes to propagate to all platforms via CI/CD; not real-time","Plugin marketplace approval (Claude Code) adds additional latency; Codex/Gemini CLI updates may lag behind GitHub releases","No rollback mechanism if sync introduces breaking changes; manual intervention required to revert across all platforms"],"requires":["GitHub repository with GitHub Actions enabled","CI/CD workflow scripts that parse SKILL.md and generate platform-specific configs","Write access to plugin repositories/registries for each platform","Platform-specific API credentials (if automated deployment is required)"],"input_types":["SKILL.md configuration file","platform-specific configuration templates"],"output_types":["platform-specific configuration files (plugin.json, Codex config, etc.)","CI/CD deployment logs","plugin marketplace submissions"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":38,"verified":false,"data_access_risk":"high","permissions":["Claude Code plugin infrastructure OR Codex CLI OR Gemini CLI","Access to SKILL.md configuration file (editable by users)","Node.js 14+ for npx skill installation method","Python 3.7+ for benchmarking/testing utilities","Node.js 14+ for npx skill installation","Bash/Zsh (macOS/Linux/WSL) OR PowerShell 5.0+ (Windows)","Write access to ~/.claude/settings.json or equivalent platform config directory","Git (for cloning repository and CI/CD workflows)","GitHub Actions (for SKILL.md sync and GitHub Pages deployment)","Claude Code, Codex, or Gemini CLI with caveman skill installed"],"failure_modes":["Intensity levels (Lite/Full/Ultra) are globally scoped—cannot apply different compression ratios to different conversation sections without manual toggling","Rule engine is linguistic-pattern-based, not semantic—may over-compress domain-specific jargon if not explicitly whitelisted in SKILL.md","No built-in context awareness for when compression should be disabled (e.g., user-facing documentation generation)","Requires explicit activation per session; no persistent state across Claude Code instances","SKILL.md sync workflow requires manual GitHub Actions trigger or CI/CD integration—no real-time propagation to installed instances","Platform-specific configuration paths differ (~/.claude/settings.json vs Codex config vs Gemini CLI)—installation scripts must be maintained per OS/platform combination","No rollback mechanism if SKILL.md update breaks existing workflows; users must manually revert configuration","Plugin marketplace approval process (Claude Code) adds latency to distribution; Codex/Gemini CLI updates may lag behind GitHub releases","Intensity levels are binary toggles per session—no mid-conversation switching without manual command invocation","No automatic intensity selection based on task type or context length; users must manually choose level","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.4327550112225126,"quality":0.43,"ecosystem":0.6000000000000001,"match_graph":0.25,"freshness":0.75,"weights":{"adoption":0.15,"quality":0.25,"ecosystem":0.1,"match_graph":0.45,"freshness":0.05}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-05-24T12:16:21.550Z","last_scraped_at":"2026-05-03T13:57:19.180Z","last_commit":"2026-05-01T00:31:28Z"},"community":{"stars":52794,"forks":2857,"weekly_downloads":null,"model_downloads":null,"model_likes":null}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=juliusbrussee--caveman","compare_url":"https://unfragile.ai/compare?artifact=juliusbrussee--caveman"}},"signature":"kvxWHLceYFuHBAKk2OUrhrBJnfGQ3AzY/WDKdJPRTLty7iNH+88ogRFVh5MvkTLyfDpVMvkpSdRoM2MbIz0rDA==","signedAt":"2026-06-21T17:27:41.466Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/juliusbrussee--caveman","artifact":"https://unfragile.ai/juliusbrussee--caveman","verify":"https://unfragile.ai/api/v1/verify?slug=juliusbrussee--caveman","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}