Skill UI

A comprehensive guide for diagnosing and reducing high memory consumption in Qdrant. This resource is crucial when dealing with reports of 'memory leak', 'out of memory', or excessive RAM growth. It details how Qdrant manages resident memory and OS page cache, and provides advanced techniques such as quantization, using float16/int8 data types, and strategically storing components on disk (like HNSW indexes or sparse vectors) to minimize the memory footprint and ensure stable, scalable operation.

This guide provides essential knowledge on resource limitations within WebContainer environments (such as memory, CPU, file system size, and process count). It is crucial for developers utilizing the WebContainers API or StackBlitz SDK to understand these constraints. Learn best practices for monitoring memory usage, optimizing mounted file systems by excluding unnecessary assets (like test or documentation folders), and managing dependencies to prevent common issues like Out-Of-Memory (OOM) crashes and slow build times.

This expert playbook details advanced techniques for converting an arbitrary memory write primitive into full Remote Code Execution (RCE). It covers targeting multiple system structures across different glibc versions, including GOT entries, __malloc_hook, __free_hook, _IO_FILE vtables, and function arrays. Essential for exploiting modern, hardened binaries and mastering advanced binary exploitation.

A comprehensive playbook detailing advanced techniques for exploiting glibc heap vulnerabilities using ptmalloc2. Covers various memory corruption primitives, including Use-After-Free, double free, and overflows. Provides version-specific guidance on bypassing modern protections like safe-linking (glibc >= 2.32) and RCE methods for post-2.34 environments.

This playbook provides advanced techniques for symbolic execution using angr, Z3, and Unicorn Engine. It covers automating CTF challenges, handling constraint solving patterns, function hooking, memory manipulation, and emulating custom virtual machines for binary analysis and key recovery.

This skill guides the design and optimization of production-grade multi-agent systems. It specializes in using LangGraph, LangChain, and DeepAgents to create structured, complex AI workflows. It covers state management, agent routing, memory integration, and tool calling, making it ideal for building robust, scalable agents like supervisors, planners, and coders.

Diagnoses full-stack failure modes in agent and LLM applications, including wrapper regressions, memory contamination, tool discipline failures, hidden repair loops, and rendering corruption. Produces severity-ranked findings and code-first fixes for production agent systems.

Design a persistent multi-agent operating system on Claude Code with kernel routing, specialist agents, slash commands, file-based memory, and scheduled automation for long-lived projects.

This skill captures and persists critical user decisions, project constraints, and architectural choices across multiple sessions and agents. It saves this context into a structured memory database (`mdbase`), ensuring that future development steps or agents can retrieve the original intent without re-deriving it from scratch. Ideal for documenting project rules, tech stacks, and design rationale.

This meta-skill automatically inspects open GitHub Pull Requests, failing CI runs, and new issues. It utilizes the GitHub API to gather critical updates and summarizes the findings into three actionable buckets (To Review, Awaiting Me, CI Red). Finally, it persists follow-up tasks to long-term memory, helping developers efficiently prioritize their daily workflow.

A meta-skill designed to monitor long-running processes, such as model fine-tunes or CI builds, executed within a detached tmux session. It captures output, automatically uses a sub-agent to diagnose failures, proposes a fix, and records the complete debugging process to memory. Ideal for intermittent failure debugging and continuous integration pipelines.

Atlas Ledger is a critical companion designed to provide lasting 'memory' to project contracts. It automatically processes detected deviations or 'drift' (e.g., from audits or reviews) and distills these instances into formal, actionable contract clauses (WHEN/DON'T/INSTEAD). These clauses are written to a persistent project memory file, Atlas.md. This mechanism ensures that lessons learned from failure are systematically captured, allowing future contract cycles to incorporate these rules, thereby strengthening overall project governance and reliability.