Skills, tools
andAI pulse.

This comprehensive system optimizes AI agent performance, offering production-ready agents, skills, hooks, and MCP configurations. It's ideal for building robust, continuously learning AI agents across various platforms like Claude Code, Codex, and Gemini, ensuring high-performance, secure, and scalable AI solutions.

This skill transforms LLM responses and memory files into a highly compressed 'caveman' dialect, cutting token usage by up to 75% without sacrificing technical detail. It's ideal for developers seeking to optimize costs, accelerate response times, and improve clarity in AI-assisted coding environments.

This skill provides a comprehensive diagnostic and optimization tool for Agent Skills. It goes beyond typical static checks by analyzing real session transcripts to measure trigger rates, user satisfaction, workflow completion, and undertrigger gaps. It combines this with research-backed static analysis across 14 checks (e.g., YAML safety, CSO compliance, info position). The tool generates a prioritized report with P0/P1/P2 fixes, scoring skills on a 5-point composite scale across 6 dimensions (Trigger Rate, User Reaction, Workflow Completion, Static Quality, Undertrigger, Token Economics) an

This repository offers a curated collection of 40 plug-and-play product management skills for AI agents, encompassing phase, foundation, and utility skills. It provides templates and workflows to ensure consistent, professional PM outputs throughout the entire product lifecycle, from discovery to iteration.

AgentShield provides a comprehensive security health check for AI agent skills, MCP servers, and plugins by aggregating 13 independent scanning engines. It's ideal for developers and users who need to quickly verify the safety and integrity of third-party AI artifacts before installation, ensuring protection against backdoors, data theft, and prompt injection.

This kit provides persistent memory for Claude Code agents, solving the problem of Claude forgetting context between sessions. It features a four-layer memory architecture (daily logs, hot cache, knowledge concepts, rules) and an agent-driven promotion pipeline where patterns observed in conversations are audited and proposed for inclusion in the knowledge base or rules. The "user only talks, agent writes" principle ensures seamless operation. It supports multi-project isolation, offers silent safety hooks, and direct control via slash operators, all stored in plain text files for easy versio