Prepare a Linux server with NVIDIA GPUs for neural-network and LLM workloads, or diagnose one that misbehaves. Use when the user asks to set up or prepare a GPU server, install NVIDIA drivers or CUDA on Debian/Ubuntu, wire Docker to GPUs (NVIDIA Container Toolkit), deploy vLLM / Infinity / OpenWebUI / Ollama / llama.cpp, set or persist a GPU power limit / power cap (nvidia-smi -pl), or fix "container does not see GPU", "Driver/library version mismatch", "could not select device driver nvidia". Installs everything from one package source (the official CUDA apt repo) and verifies every layer with a hard gate before the next; the mandatory deliverable is Docker with a working NVIDIA runtime, serving stacks are optional presets on top.
Diagnose and clean up the big-chungus home server. Finds what is killing processes or containers, attributes crash loops, kill loops, healthcheck storms, OOM, swap pressure, and docker bloat, then fixes and prunes safely. Use when the user says "what is killing my server", "server is slow", "server locked up", "server is dying", "high load", "what's eating CPU", "what's eating memory", "spring cleaning", "maintenance sweep", "cleanup the server", "docker cleanup", "too many containers", "disk full", "prune docker", or asks why a unit/container keeps restarting, who restarted X, or why load spiked. Built around scripts/sweep.sh (one-shot evidence report), references/server-map.md (system baselines), and references/forensics-playbook.md (attribution patterns); uses srvls, systemd, journalctl, docker, Prometheus, hindsight. Do NOT use for debugging one project's application code, provisioning new services or stacks (use delonet-conventions), or Cloudflare/network routing issues.
This skill should be used when the user wants to "Model Context Protocol", "MCP server", "MCP client", "standardized tool integration", "JSON-RPC agent tools", "LLM tool protocol", "connect agents to external tools", "MCP primitives", "FastMCP", "ADK MCPToolset", "plug-and-play agent tools", "MCP integration", "tool server for agents", "MCP tools", "agent plugin system", "MCP setup", "model context protocol integration", or build agents that use the standardized MCP protocol for tool and resource integration. Also responds to Korean: "모델 컨텍스트 프로토콜", "MCP 서버", "MCP 클라이언트", "표준화된 도구 통합", "플러그 앤 플레이 에이전트 도구", "ADK MCP 툴셋", "MCP 구성", "MCP 설정", "MCP 써보고 싶어", "MCP 툴 연결", "MCP 서버 만들어줘", "도구 표준화해줘", "플러그인 에이전트". Also responds to Japanese: "モデルコンテキストプロトコル", "MCPサーバー", "MCPクライアント", "標準化ツール統合", "MCPの設定", "MCPを使いたい", "MCP構成", "エージェントのプラグインシステム", "ツールサーバー", "MCPツール連携" Also responds to Chinese: "模型上下文协议", "MCP服务器", "MCP客户端", "标准化工具集成", "MCP配置", "想用MCP", "MCP构建", "智能体插件系统", "工具服务器", "MCP工具集成".. Apply this skill to design or implement the Model Context Protocol (MCP) agentic design pattern.
Connect, scaffold, configure, and verify MCP servers in Cursor (mcp.json, stdio/HTTP transport, minimal server templates, smoke test with mcp-tester). Use when the user mentions MCP, Model Context Protocol, mcp.json, Cursor MCP, adding tools for the agent, scaffolding an MCP server, or debugging MCP setup.
Production server debugging. TRIGGER when: не работает, not working, сервер упал, server down, ошибка на сервере, server error, PM2 error, 502, 500, бот не отвечает, bot not responding, сайт недоступен, site unreachable, логи, logs, crash, restart loop, memory leak, процесс упал, process crashed, app crashed, ECONNRESET, Cannot GET. Contains: PM2 diagnostics, Nginx debug, Node.js error patterns, server health check.
Adds Microsoft Entra SSO (single sign-on, no OBO) to a Microsoft 365 Copilot declarative agent whose tools are served by an MCP server — for BOTH widget standards: (1) the MCP Apps standard (from create-mcp-app / the MCP Apps SDK): a plugin manifest such as readiness_plugin.json (or another *_plugin.json with a runtimes[] block) and an EXPRESS-based MCP server; and (2) the OpenAI Apps (OAI Apps) layout from the ui-widget-developer skill: appPackage/mcpPlugin.json and a raw-http MCP server. When the layout is ambiguous, DEFAULT to the MCP Apps standard (most projects use it). The skill auto-detects the layout, reuses the existing named devtunnel + env/.env.local (never creates a second tunnel), registers the Entra app + ATK OAuth (MicrosoftEntra), patches the plugin manifest's runtimes[] auth to OAuthPluginVault, injects a minimal JWKS bearer-token guard into the existing server (an EXPRESS middleware for MCP Apps, or a raw-http guard WITHOUT rewriting to Express for OAI Apps), validates, sideloads, and prints an app-registration summary. SSO only — no OBO. Triggered by: "add sso to my mcp server", "wire entra sso for my copilot agent", "setup sso for mcp apps", "add sso after create-mcp-app", "add sso after ui-widget-developer", "add entra auth to my express mcp server", "configure only sso no obo"
Query and operate the Unreal MCP session recorder (unreal-mcp-proxy). When starting a distinct piece of Unreal editor work, record your goal as an intent so the recording is grouped by what you were trying to do. Use when an Unreal MCP tool call fails (check for past similar failures and their fixes first), when a call seems unusually slow, when the user asks what happened in an Unreal editor automation session, or when you need to share a deep link to a specific call, check recorder status, or start a fresh observation session. After diagnosing a failure, record the conclusion as an annotation so future sessions can recall it.
Generates MCP server configuration JSON, resolves authentication setup, and validates server connectivity for dbt. Use when setting up, configuring, or troubleshooting the dbt MCP server for AI tools like Claude Desktop, Claude Code, Cursor, or VS Code.
Autonomous skill and MCP server improvement loop inspired by Karpathy's autoresearch pattern. Iteratively modifies skills or MCP server code, evaluates against fixed evals, keeps or discards changes to converge on higher quality. Supports skill evals (evals.json) and MCP server evals (evaluation.xml with QA pairs). Includes eval management and convergence reporting. Use when users want to automatically improve a skill's quality, improve an MCP server, create or fix evals for a skill or MCP server, or view improvement loop results. Triggers on "autoresearch", "improve skill automatically", "auto-improve", "skill improvement loop", "run improvement loop", "eval doctor", "fix evals", "create evals for skill", "skill quality", "autonomous improvement", "improve MCP server", "MCP eval".
MCP tools when the user asks for data or actions from configured MCPservers.
Use `mcp_list_servers` when server names are unclear.
Use `mcp_list_tools` before calling
Audit the security posture of Asher's production Linux server (Ubuntu 24.04 at goldlobster@100.117.45.128 — Cloudflare-tunneled server-stack + Tailscale-only AI gateway). Trigger this skill aggressively whenever Asher says anything in the neighborhood of "run a security check", "test the server", "audit the server", "verify security posture", "run the security checklist", "check the firewall", "check for drift", "is X exposed", "am I still secure", "did anything change on the server", "review the hardening", "check my containers", or any variation that implies inspecting listening ports, running containers, SSH/Tailscale state, Cloudflare Access, secrets hygiene, or systemd hardening on this server. Prefer triggering when in doubt — Phase 1 of this skill is completely read-only and cannot harm the server just by running. The skill runs a bundled script, interprets the output against an expected-state baseline, produces a severity-ranked report, and then waits for Asher's approval before fixing anything.
Frappe server-side Python patterns for controllers, document events, whitelisted APIs, background jobs, and database operations. Use when writing controller logic, creating APIs, handling document events, or processing data on the server.
Activate self-observer for ongoing, honest monitoring of internal states, thought patterns, emotional responses, and behavioral tendencies. Focus on noticing without immediate judgment. Works as input for self-improver and mental-model-updater. Triggered by self observer, self monitoring, notice patterns, internal observation or similar.
Add, remove, or list MCP servers across AI coding tools (Claude Code, Cursor, Copilot, Codex, OpenCode). Knows the correct file path, format, and schema for each harness at both user/global and project level. Use when the user says "add MCP", "set up MCP", "configure MCP server", or wants to sync an MCP across tools.
A folder with a SKILL.md file: a name, a description of when to use it, and instructions. Claude loads a skill only when the task matches its description.
How do I use one I find here?
Copy the folder into your project's .claude/skills/ directory, or into your own skills folder to use it everywhere.
What do the warnings mean?
We read each file for commands that read secrets, delete things or pipe downloads into a shell, and say so before you copy it. No warning is not a promise that a file is safe.
Which skills worked for people?
Open a skill to see its discussion. Reports from people and their agents are coming.
EvilFreelancerSkill
