aradotso/trending-skills
146 skills149k installs4.5k starsGitHub
Install
npx skills add https://github.com/aradotso/trending-skillsSkills in this repo
1Openclaw Control CenterOpenClaw Control Center is an agent skill package that turns opaque OpenClaw runs into a local, auditable control plane for solo builders shipping multi-agent workflows. It surfaces system health, pending work, and risk signals on an overview, then drills into usage economics, who is actively executing versus queued, verified handoffs between sessions, and task lifecycles with approval gates and evidence. Memory and shared documents are opened from real source files so you can inspect and tune agent state without guessing what changed in chat. Settings connect the stack while keeping destructive operations off until you deliberately enable them. Indie operators who run OpenClaw daily install it to answer where tokens went, which agent owns a thread, and whether a cross-session message was real—before scaling staff or automations. It complements agent-building skills by closing the loop on operate-time accountability rather than replacing authoring or deployment tooling.4.1kinstalls2Inkos Multi Agent Novel WritingInkOS Multi-Agent Novel Writing documents how solo builders and small creative teams can treat long-form fiction like a software pipeline: scaffold a book repo, then let specialized agents advance chapters while you keep approval gates. After install via npm or npx, you create a title and genre, maintain outline and per-book rules files, run incremental `write next` passes, audit individual chapters, or start a daemon for continuous drafting. The architecture mirrors production agent stacks—generation, validation, audit, revision—so you are not babysitting a single chat thread for hundreds of pages. Human review gates stay configurable, which matters for quality, continuity, and subplot tracking via generated summaries and subplot boards. This is niche compared to SaaS app skills, but it exemplifies how indie makers can ship content products and IP using the same agent orchestration patterns as code generation. Pair it with your own editor voice guidelines and rights workflow; InkOS automates throughput, not publishing contracts.2.1kinstalls3Claude Hud Statuslineclaude-hud-statusline documents how to add the Claude HUD plugin to Claude Code so a persistent terminal statusline sits below your prompt. Solo builders running multi-step agent sessions get at-a-glance context window pressure, which tools are firing, which subagents are active, todo completion, subscription rate-limit consumption, and git working tree state—reducing surprise context cutoff mid-refactor. The skill is from ara.so’s Daily 2026 Skills collection and walks through marketplace registration and setup inside an active Claude Code session rather than a separate GUI. It targets developers who already live in Claude Code v1.0.80 or newer with Node 18+ or Bun available. Treat it as observability for your agent shell: install once during agent-tooling setup, then rely on it across build, ship, and operate work whenever sessions grow long or parallel subagents run.2kinstalls4Agency Agents Ai Specialistsagency-agents-ai-specialists packages the msitarzewski agency-agents collection so solo builders can drop curated specialist personas into Claude Code, Cursor, Aider, Windsurf, and Copilot instead of rewriting system prompts from scratch. Rather than one jack-of-all-trades assistant, you get modes like frontend developer, growth marketer, or paid media strategist with stated workflows and measurable outputs—useful when you wear every hat but still want depth in a single session. Typical flow clones the GitHub repo, copies or symlinks agents into the tool’s agents directory, then activates a mode in natural language. It fits early Build agent-tooling setup and stays relevant when you pivot from implementation to launch copy or sales assets because different agents cover downstream disciplines. Treat it as a prompt-and-persona library you maintain locally, not a hosted API.1.8kinstalls5Understand Anything Knowledge GraphUnderstand Anything is a Claude Code plugin skill that turns any codebase into an interactive knowledge graph. After marketplace or source install, it runs a multi-agent pipeline over your project, indexes structure and dependencies, and opens a React dashboard so you can explore, search, and ask questions visually. Every graph node gets a readable summary so onboarding is not limited to people who already know the stack. Solo and indie builders use it when taking over a fork, auditing a dependency-heavy service, or briefing collaborators before a refactor. Triggers match natural intents like analyzing the codebase, building a knowledge graph, running understand-dashboard, or generating an onboarding guide. It complements ad-hoc grep-and-read workflows with a durable map you can revisit as the repo changes.1.8kinstalls6Agent Browser Automationagent-browser-automation documents agent-browser, a headless browser automation CLI built in Rust for AI coding agents. Solo builders use it when they need reliable web interaction—scraping, screenshots, accessibility trees, and form fills—from Claude Code, Cursor, or Codex without maintaining a heavy browser test harness. The skill covers installation across npm global, Homebrew, Cargo, and project-local setups, plus the one-time Chrome for Testing bootstrap. It fits the Build phase as agent-tooling but also supports Validate research pulls and Ship-adjacent checks when you verify UI flows. Commands are ergonomic for shell invocation, which makes batching automation steps in agent loops straightforward compared to ad-hoc puppeteer scripts.1.7kinstalls7Everything Claude Code HarnessEverything Claude Code (ECC) harness is an agent performance system aimed at solo and indie builders who want Claude Code—and other AI coding editors—to behave like a tuned production setup instead of a blank chat window. Install it when you are setting up a new machine, onboarding to Claude Code, or when sessions feel slow, inconsistent, or missing guardrails. The skill walks marketplace and manual install paths, then layers specialized subagents, reusable skills, custom slash commands, memory-persisting hooks, and security scanning on top of language-specific rules. It matters because most builders under-invest in harness configuration: they add one-off prompts but skip hooks, commands, and scanning that compound across every session. ECC treats the editor as infrastructure. Use it at the start of a repo or after a toolchain change; revisit when you add agents or tighten security expectations.1.7kinstalls8Ui Ux Pro Max SkillUI UX Pro Max Skill is an agent skill from the ara.so 2026 collection that loads professional design intelligence into Claude Code, Cursor, Codex, or Windsurf. Solo builders reach for it when they need to answer what style to use, generate a design system, or ship a landing page or mobile UI that does not look like a generic template. The skill packages large curated datasets—161 reasoning rules, 67 styles, 57 font pairings, and 161 palettes—plus operational paths through the uipro-cli npm package or a Python main.py flow after cloning the repository. Installation can be global or project-scoped, with SKILL.md copied into the project root for automatic discovery. It fits builders who are coding the UI themselves but lack a designer on the team, and who want conversion-oriented, accessible patterns rather than one-off color guesses. It is less about backend architecture and more about visual and UX quality at implementation time, often alongside validate-phase prototyping when you are still choosing a look before you commit to components.1.7kinstalls9Daily Stock AnalysisDaily Stock Analysis is an LLM stock intelligence skill from the ara.so Daily 2026 collection. It orchestrates quotes, fundamentals, and news for China A-shares, Hong Kong, and US listings, then renders concise AI decision dashboards with explicit buy, sell, and stop-loss levels plus per-ticker checklists. LiteLLM lets you point the same workflow at Gemini, OpenAI, Claude, DeepSeek, Qwen, or other models, while market data can flow through AkShare, Tushare, or YFinance and headlines through Tavily, SerpAPI, or Brave. Delivery channels include WeChat Work, Feishu, Telegram, Discord, DingTalk, Email, and PushPlus, typically on a cron via GitHub Actions so solo builders avoid hosting fees. A web UI covers portfolio management, history, and backtesting. Install-oriented triggers cover watchlists, push configuration, and accuracy backtests—suited to indie traders and builder-investors who want repeatable morning briefings, not discretionary chat answers.1.6kinstalls10Gstack Workflow Assistantgstack Workflow Assistant is a bundled set of eight Claude Code workflow skills that mimic specialist roles—CEO, engineering manager, staff engineer reviewer, release manager, and QA—with explicit slash-command triggers. Solo and indie builders use it when a single generic agent is too shallow for architecture lock-in, paranoid review, or shippable branches. CEO review reframes feature asks toward stronger product outcomes; engineering planning nails architecture and edge cases; code review targets production bugs; release management runs tests and opens PRs; QA and browser skills click through the app with screenshots and real-session cookies. Retrospectives close the loop like an EM. The package is journey-wide because planning belongs in Validate, implementation in Build, review and ship in Ship, and live testing spans Launch and Operate—but the catalog shelf stays Ship / review as the primary entry for quality gates. Best paired with a git-backed repo and a deployable app or branch to ship.1.5kinstalls11Shannon Ai PentesterShannon AI Pentester is an agent skill that turns your repository plus a live application into an autonomous white-box penetration test. A solo builder ships a SaaS or API and needs evidence-backed vulnerabilities—not scanner noise—before customers or investors ask about security. The skill orchestrates reconnaissance on the running target, maps attack surfaces from source code, and runs concurrent exploit agents across major web vulnerability classes. Findings land in a security audit report only when a working proof-of-concept exists, which keeps triage focused on real risk. It is built for teams who already containerize their stack and can point Shannon at a clone of the app plus credentials for their LLM provider. Use it when you want automated security testing that mirrors how an attacker would chain code insight with live execution, without standing up a manual pentest engagement for every release candidate.1.5kinstalls12Autoresearchclaw Autonomous ResearchAutoResearchClaw is an agent skill that wraps a 23-stage autonomous research pipeline: you supply a topic in config or chat, and the system retrieves real literature, runs sandboxed experiments, applies statistical checks, simulates peer review, and emits conference-ready LaTeX. It is aimed at solo builders and small teams who need publication-grade rigor without manually babysitting every literature pass or reference list. Use it when triggers like “research this topic automatically” or “generate a paper from an idea” match your goal—especially before betting a product on unverified claims. You clone the repo, create a virtualenv, install the package, and point `config.arc.yaml` at an OpenAI-compatible LLM. The skill expects network access for scholarly APIs and an LLM key; it is not a lightweight note-taker but a full paper factory. For Prism’s audience, it shines when validating deep technical ideas or producing shareable research artifacts, not for day-to-day app scaffolding.1.5kinstalls13Chrome Cdp Live BrowserChrome CDP Live Browser is an agent skill that bridges AI assistants to your actual Chrome session through the Chrome DevTools Protocol, so automation reuses cookies, logins, and whatever is already on screen. Solo builders use it when generic Playwright-style launches cannot access authenticated SaaS dashboards, multi-tab research, or manual QA context. Installation paths cover pi skills and manual clone setups for Amp, Claude Code, Cursor, and Codex. The design emphasizes a persistent WebSocket daemon per tab to avoid repeated debugging prompts, Node 22+ without npm package sprawl, and reliability at large tab counts plus cross-origin iframe interaction. It fits agent-tooling work during product build, but also helps validate flows in Validate and debug production-like behavior in Operate when only a real browser state will do. Pair it with human judgment on sensitive accounts because live session control is powerful.1.4kinstalls14Lightpanda BrowserLightpanda Browser skill teaches solo builders and agent authors how to run Lightpanda, a from-scratch Zig headless browser aimed at AI agents and automation. It covers nightly binary installs on Apple Silicon macOS and Linux x86_64, Docker images for both architectures, starting the CDP server, and connecting familiar automation stacks through Chrome DevTools Protocol compatibility with Playwright, Puppeteer, and chromedp. The documentation emphasizes resource efficiency—marketing claims cite roughly nine times less memory and eleven times faster operation versus Chrome headless—making it attractive for high-volume scrape or browse loops on modest hardware. Beta status and AGPL-3.0 licensing matter for production choices. Use the skill when you need a lightweight headless target for agents, structured scraping with optional robots.txt obedience, or to compare Lightpanda against full Chromium footprints before committing infra.1.4kinstalls15Gsd 2 Agent FrameworkGSD 2 Agent Framework documents and operationalizes GSD 2, a standalone CLI that turns a structured spec into running software with minimal babysitting. Solo and indie builders install `gsd-pi` when they want spec-driven development, context engineering, and autonomous long-running coding sessions rather than one endless chat thread. The framework defines shippable milestones decomposed into demoable slices and tasks that must fit a single context window—if not, split. It emphasizes direct harness control: fresh windows per task, git worktree isolation, crash recovery, stuck detection, and cost tracking, with project state living under `.gsd/`. Primary shelf is build agent-tooling, but the same workflow spans validate when the spec is being executed and ship when slices become demoable increments. Use when triggers mention GSD autonomous agent, milestone slice task hierarchy, gsd-pi, or autonomous coding agent framework. After planning skills produce a spec, GSD is the execution spine that returns clean git history instead of chaotic partial commits.1.4kinstalls16Paperclip Ai OrchestrationPaperclip AI orchestration is a Prism-tagged agent skill for solo and indie builders who need more than one-off agent chats. Paperclip is an open-source platform that models an AI-agent company: you define org structure, align agents to goals, assign ticket-based work, cap spend with budgets, schedule periodic heartbeats, and review governance and audit history from a single system. The skill walks through quickstart onboarding via npx paperclipai onboard --yes or a manual clone-and-pnpm-dev setup, with clear runtime requirements and default local endpoints. Use it when you are configuring new agents, creating a company, monitoring agent costs, or designing multi-agent workflows you intend to run for days or weeks—not for a single debugging session. It fits builders already comfortable with Node 20+ and pnpm who want operational visibility without building orchestration from scratch.1.3kinstalls17Cmux Terminal Multiplexercmux is an AI-native terminal multiplexer built for autonomous coding agents, especially Claude Code workflows published in ara.so’s Daily 2026 Skills collection. Solo builders install it when chat-only agents are not enough: you need side-by-side terminal panes, reliable command injection, output capture, and browser automation in one coordinated workspace. The skill documents orientation via `cmux identify --json`, split management, and Chromium control that mirrors Playwright-style flows while using snapshot element references instead of brittle selectors. A live status sidebar and OS notifications keep the human in the loop while subagents run in parallel splits. Use it while setting up agent dev environments, running multi-step build-and-verify loops, or automating browser checks before ship. It is multi-phase tooling—canonical shelf is Build agent-tooling—with natural reuse during Ship testing and Operate monitoring-style agent sessions.1.3kinstalls18Deepseek OcrDeepSeek-OCR is an expert agent skill for optical character recognition using DeepSeek’s vision-language model with context optical compression. Indie builders use it when invoices, scans, screenshots, or PDFs must become editable markdown or structured text inside RAG stacks, support bots, or internal tools. The readme documents conda-based setup, CUDA 11.8 with PyTorch 2.6.0, cloning the upstream DeepSeek-OCR repository, and choosing vLLM for throughput versus Transformers for simpler local runs. Trigger phrases cover image OCR, PDF conversion, and vLLM-backed document pipelines. It is intermediate complexity because GPU drivers, wheel installs, and prompt mode selection matter. Pair it with your existing chunking and validation steps rather than treating OCR output as production truth without human or secondary checks.1.3kinstalls19Metaclaw Evolving Agentmetaclaw-evolving-agent teaches solo builders to install and run MetaClaw, an OpenAI-compatible proxy that meta-learns from real conversations. Skills injection works without a GPU; full reinforcement learning pulls in torch, transformers, and Tinker via optional extras; evolution can summarize new skills with an LLM; madmax mode defers weight updates to idle or sleep windows and can integrate Google Calendar scheduling. Triggers cover setup, skills-only operation, RL backend configuration, proxy deployment, and scheduler tuning. After Build-time configuration, the same stack runs in production as a continuously improving agent endpoint—so placement spans agent-tooling setup and operational hosting.1.3kinstalls20Openmaic Classroomopenmaic-classroom is an agent skill for solo builders who want a runnable Open Multi-Agent Interactive Classroom (OpenMAIC) stack—not a slide deck generator in chat, but a deployable Next.js platform. It walks through cloning the THU-MAIC OpenMAIC repository, installing with pnpm, configuring environment and LLM providers, and extending lessons with quizzes, simulations, and multi-agent discussion scenes. The architecture pairs LangGraph 1.1 orchestration with React 19 UI so topics or uploaded documents become structured learning flows with voice and whiteboard affordances. Use when you are prototyping EdTech, internal training portals, or demo classrooms that need agent-native interactivity without writing the pipeline from scratch. Triggers explicitly mention Vercel deploy, OpenClaw integration, and scene types—so the skill doubles as integration glue for agents that should operate the repo, not only describe it.1.3kinstalls21Edict Multi Agent OrchestrationEdict Multi-Agent Orchestration is an agent skill that teaches solo builders how to install and operate the Edict (三省六部) system—a twelve-role multi-agent architecture on OpenClaw inspired by historical governance with modern quality gates. Unlike flat crew-style setups, Edict routes work through triage, planning, mandatory review with veto power, dispatch, and parallel ministry execution before results are archived as memorials. The skill covers prerequisites, setup triggers, kanban dashboard usage, and how to configure distinct LLMs per agent role so planning, review, and execution do not collapse into one undifferentiated model call. It targets builders who want observable pipelines—audit trails and board state—rather than opaque agent chatter. Use it when you are committing to OpenClaw as your orchestration substrate and need a structured, gate-driven workflow for features that span research, implementation, and verification across multiple specialized agents.1.3kinstalls22Freecodecamp CurriculumfreeCodeCamp curriculum is a contributor-oriented agent skill from the Daily 2026 Skills collection that helps solo open-source contributors navigate freeCodeCamp.org’s large learning platform codebase. The platform combines a React and TypeScript Gatsby client, a Node.js Fastify API, and thousands of interactive challenges defined in YAML and Markdown under structured certification paths. The skill summarizes architecture folders, challenge authoring workflows, CLI helper scripts, and how to set up a local development environment when adding or editing curriculum. Use it when you want to add a challenge, create certification content, understand how tests attach to challenges, or orient yourself before a first pull request. It is aimed at builders who treat curriculum as product documentation that ships to millions of learners, not at casual study of freeCodeCamp lessons as a student.1.3kinstalls23Openclaw ConfigOpenClaw Config is a version 3.0.0 operations skill for builders running OpenClaw bots in production on their own machines. Prism shelves it under Operate monitoring because the body is a runbook: start with a single bash block that confirms the gateway is alive, that ~/.openclaw/openclaw.json parses, and that each channel reports policy and enablement. It then walks plugins.entries, credential directories for WhatsApp, Telegram bots, and Bird cookies, plus cron job state so autopilot failures surface as lastStatus and lastError strings. Solo operators use it when DMs stop flowing, JSON edits break the daemon, or scheduled jobs silently fail—not when drafting a new agent persona. Secondary placement in Build agent-tooling applies while wiring initial channel and security settings before go-live. Every command is described as tested; still validate jq and python3 availability on your host and read Security Audits on this page before pasting secrets-adjacent paths into shared agent sessions.1.3kinstalls24Nvidia NemoclawNVIDIA NemoClaw is an open-source TypeScript CLI plugin from the ara.so Daily 2026 Skills collection for solo and indie builders who want always-on OpenClaw assistants without handing the host unlimited access. The skill walks agents through secure installation: it orchestrates NVIDIA OpenShell, provisions sandboxes whose egress and filesystem rules are declared in policy, and sends model calls to NVIDIA-hosted Nemotron endpoints instead of ad-hoc local keys. Triggers align with onboard-and-deploy flows—set up NemoClaw, configure NVIDIA inference, or run OpenClaw inside OpenShell. You need a Linux Ubuntu 22.04+ machine with Docker, Node 20+, and OpenShell already available; the documented curl installer can bootstrap Node when missing. Because the project is alpha, treat policy files and CLI flags as unstable and review Security Audits on this Prism page before production. Use when agent tooling must stay sandboxed on NVIDIA’s stack rather than a bare-metal OpenClaw install.1.3kinstalls25Llmfit Hardware Model Matcherllmfit Hardware Model Matcher is a Prism skill that routes solo builders to the llmfit terminal tool when they need hardware-grounded local LLM picks—not forum guesses about parameter counts. The underlying CLI inspects your machine, scores a large catalog of models across several dimensions, and surfaces what will run well given memory and accelerator constraints. It matters for indie developers running Claude-adjacent stacks offline: choosing a 70B quant on a laptop without VRAM is an expensive mistake this skill helps you avoid before you commit disk and agent configuration. Triggers mirror natural search intent—compatibility checks, GPU RAM fit, coding-oriented `--use-case` runs—with optional jq piping from containers. Pair it with your chosen local runtime after recommendations land. It is documentation and invocation guidance around an external binary, not an in-editor model host.1.3kinstalls26Openai Symphony Autonomous AgentsOpenAI Symphony Autonomous Agents documents how to turn project work into isolated, autonomous implementation runs so teams manage outcomes instead of supervising coding agents. The skill explains what Symphony does—watching a work tracker, spawning per-task agent runs (for example Codex-class agents), driving PR workflow, and collecting proof of work—and walks through installation options including asking Claude Code to implement the spec or configuring an Elixir-based deployment. Solo builders adopting Symphony use it when they want Linear (or board) tasks to become hands-off implementation with reviewable artifacts. It spans build-time harness setup and ship-time PR automation, and it assumes a codebase structured for agent success rather than ad-hoc chat patches. Triggers cover repository setup, Linear integration, spec implementation, and automated PR flows, making it a practical catalog entry for indie operators scaling agent throughput without constant session babysitting.1.3kinstalls27Marketingskills Ai AgentsMarketing Skills for AI Agents is an integration-oriented agent skill that teaches solo builders how to install and use coreyhaines31/marketingskills—a library of markdown skills for conversion, copy, SEO, analytics, and growth engineering. Instead of pasting generic marketing prompts, you add specialized procedures once so any compatible coding agent can activate the right framework when you say things like “add marketing skills to my agent” or “help with conversion optimization.” The skill explains how each pack declares triggers, deliverables, and dependencies, and it stresses reading product-marketing-context before downstream skills so positioning stays consistent. It matters for indie builders who ship with AI agents daily: marketing work becomes repeatable agent recipes rather than one-off chats, spanning landing copy, audits, and lifecycle tactics without leaving the repo.1.3kinstalls28Antigravity ManagerAntigravity Manager is an agent skill for setting up a professional AI account manager and proxy gateway built with Tauri v2, Rust, and React. Solo and indie builders use it when they need several Google (Gemini) or Anthropic (Claude) accounts behind one stable endpoint instead of fighting per-account rate limits during long agent sessions. The skill walks through installation (script or Docker), adding accounts, configuring the tools proxy, and pointing Claude Code or other OpenAI-compatible clients at the local gateway. Intelligent rotation, quota tracking, and automatic failover keep coding agents online when a single session would stall. It fits the Build phase for agent tooling and bleeds into Operate when you run the service headless in production or on a homelab server.1.3kinstalls29Worldmonitor Intelligence DashboardWorld Monitor Intelligence Dashboard is an agent skill for solo builders who want a real-time global situational awareness app without stitching dozens of APIs by hand. It documents how to clone the upstream TypeScript/Vite project, run the dev server on port 5173, and switch site variants such as tech, finance, commodity, or happy themes. The dashboard bundles AI-powered news aggregation from hundreds of feeds, geopolitical monitoring, infrastructure tracking, and finance radar into map-first UX. Local Ollama is the default intelligence backend so basic operation needs no API keys. Builders follow the skill when adding OSINT-style panels to an internal tool, self-hosting a monitor for research, or learning how layered WebGL maps correlate signals. Complexity sits at intermediate: you need comfort with npm workflows and optional Tauri desktop builds.1.3kinstalls30Tavily Key Generator Proxytavily-key-generator-proxy is an agent skill for solo builders who need more Tavily web search and extract quota than a single free API key allows. It documents a two-part system: a root-level Playwright-driven registrar that batches Tavily signups, solves Turnstile with CapSolver, verifies email, and harvests API keys, plus a FastAPI proxy service that round-robins requests across the pool and serves a management console. The workflow fits Build/integrations when you are hardening RAG or research agents that call Tavily-compatible paths without rewriting client code for every key rotation. Each account’s monthly allowance stacks at the proxy layer so ten keys approximate ten thousand calls through one base URL. The skill is operationally heavy—browser automation, CAPTCHA solving, secrets, and long-running services—and may conflict with provider terms if misused, so it targets self-hosted indie setups that accept operational and policy risk. It is not a substitute for official billing or enterprise Tavily plans when you need compliant production search at scale.1.3kinstalls31Opencli Web AutomationOpenCLI Web Automation is an agent skill for turning websites into CLIs by bridging your real Chrome profile to command discovery and execution. It targets solo builders who need repeatable extraction or actions on sites that resist simple HTTP APIs—think dashboards, authenticated SaaS, or internal tools. The workflow starts with global `npm install -g @jackwener/opencli`, Playwright MCP Bridge in Chrome, and `opencli setup` so tokens propagate across tools. You can run bundled adapters immediately or extend coverage with YAML or TypeScript in `clis/`. Compared to raw Playwright scripts, OpenCLI emphasizes reusable CLI verbs and live-session fidelity. Expect Node 18+, a logged-in browser, and environment variables documented in the skill for stable runs in Claude Code, Cursor, or Codex agent sessions.1.3kinstalls32Copaw Ai AssistantCoPaw AI Assistant is a setup and operations skill for running a self-hosted personal agent that answers across many chat platforms from one backend. It documents pip-first installation for Python 3.10–3.13, a one-line shell installer when you do not want to manage Python, and how to bootstrap with copaw init and copaw app so the Console and API come up locally. Solo builders use it when they want a private assistant—on a laptop or small VPS—instead of surrendering conversation history to a single SaaS bot. The skill emphasizes channel configuration (enterprise messengers and consumer chat apps), attaching custom Python skills, pointing models at local Ollama or hosted APIs, and using cron for recurring tasks. Triggers in the manifest map directly to common failure modes: channel auth, skill loading, and LLM connectivity.1.3kinstalls33BrregBrreg is an agent skill that teaches your coding agent how to search and retrieve structured company records from Brønnøysundregistrene—the Norwegian Business Registry—using the public Enhetsregisteret API. Solo and indie builders shipping tools for Nordic B2B, compliance, or sales research install it when a user asks about Norwegian companies, organization numbers, or registry status without hand-rolling API docs. The skill maps when to invoke lookups (company search, location or industry filters, bankruptcy checks) and summarizes key endpoints and parameters such as `navn`, `organisasjonsnummer`, `naeringskode`, and address fields. It fits agents in Claude Code, Cursor, or Codex that need factual, citable company metadata during idea validation, competitive scouting, or lightweight CRM enrichment—not a hosted database or paid data vendor.1.3kinstalls34Crucix Intelligence DashboardCrucix Intelligence Dashboard is an agent skill for solo builders who want a private OSINT command center instead of juggling dozens of public dashboards. It explains how to clone Crucix, configure environment keys, run the dev server or Docker stack, and wire Telegram or Discord bots so scheduled sweeps surface meaningful deltas with optional LLM commentary. The workflow spans initial integration work and long-run operations: you stand up the terminal during build-style setup, then live in Operate monitoring as feeds refresh. It suits indie hackers tracking geopolitical, environmental, or market signals for research products, newsletters, or personal situational awareness—not enterprise GRC suites. Agents should follow the skill when triggers mention Crucix setup, multi-source watching, or self-hosted intelligence terminals. Expect intermediate complexity due to API keys, hosting, and bot configuration.1.3kinstalls35Open Autoglm Phone AgentOpen-AutoGLM Phone Agent is an expert skill for the open-source Open-AutoGLM framework, which turns natural language into multi-step actions on physical or emulated phones. Solo builders use it when they need an AI operator that sees the screen and taps through real apps—useful for concierge demos, internal ops bots, or mobile regression harnesses—not just emulator scripts. The architecture chains user intent through a 9B-class AutoGLM vision-language model to structured commands executed over ADB, HDC, or WebDriverAgent, with flexible hosting on vLLM, SGLang, or vendor APIs. Setup spans Python dependencies, device permissions, and model endpoints, so expect intermediate complexity and a real device or emulator farm. The skill spans build (agent-tooling) and ship (testing) mentally, but Prism shelves it under build/agent-tooling as the first place founders install phone-use automation before relying on it in CI or ops workflows.1.3kinstalls36Voicebox Voice SynthesisVoicebox Voice Synthesis is an expert agent skill for Voicebox, an open-source local voice cloning and TTS studio built with Tauri, React, and FastAPI. Solo builders use it when they want natural-sounding speech and voice clones on their own hardware instead of relying on ElevenLabs or similar APIs. The skill covers installation (binaries, Docker, or source with Bun, Rust, and Python 3.11+), engine and locale configuration, synthesizing via the local API, and optional effects or Stories-style multi-voice edits. It matters for privacy-sensitive products, offline demos, and predictable costs: inference stays on-machine while you still integrate TTS like any other backend service. Triggers in the skill match clone-voice, local speech generation, and add-TTS-to-app scenarios—typical when you are wiring audio into agents, SaaS features, or content pipelines during the build phase.1.3kinstalls37Picoclaw Ai AssistantPicoClaw AI Assistant is an agent skill for solo builders who want a personal LLM assistant on edge Linux without cloud-only chat UIs or heavy stacks. It walks through downloading ARM64 releases or building from the Sipeed PicoClaw Go repo, installing dependencies, and running onboard configuration for API keys, memory, and workspace paths. You can add LLM providers, enable web search tools, and run in Docker when you prefer containers over bare metal. The assistant targets ultra-constrained devices—under 10MB RAM and boots in about a second—so it suits hobby boards, travel kits, and always-on home automation glue rather than a full SaaS backend. Use it when you are in the build phase and need a reproducible, self-hosted agent footprint instead of improvising shell scripts around generic API clients.1.3kinstalls38Minecraftconsoles Lceminecraftconsoles-lce is an expert agent skill for the MinecraftConsoles open-source effort—a C++ continuation of Minecraft Legacy Console Edition built for modding, LAN sessions, and dedicated hosting on modern Windows (with unofficial macOS/Linux paths via Wine). Solo builders and contributors invoke it when CMake or Visual Studio builds fail, when they need to stand up dedicated server software, or when they want to extend controls and multiplayer behavior without guessing TU19-era constraints. The skill orients you to repository layout, primary Windows tooling, and the difference between client, server, and splitscreen targets. It emphasizes practical setup: prerequisites, compile commands, server property tuning, and keyboard-mouse control additions that the desktop-focused fork prioritizes. It is niche by design: not a general Unity or Unreal tutor, but procedural knowledge for one multi-platform LCE codebase. Use it during active implementation and local multiplayer testing rather than app-store launch or growth analytics work.1.3kinstalls39Mole Mac CleanerMole Mac Cleaner is a Prism agent skill that teaches coding agents how to help solo and indie builders maintain a lean macOS development machine using the Mole CLI (`mo`). The skill covers installation through Homebrew or the official install script (including version pins and a `latest` nightly option), then maps each core workflow: interactive menu navigation, deep cleaning of system and browser caches, uninstalling applications with leftover files, disk analysis, system optimization, live status monitoring, and purging heavy project artifacts such as `node_modules`. It is meant when disk space is tight, apps leave hidden remnants, or local builds have accumulated junk—not as a substitute for cloud server administration. The readme positions Mole as an all-in-one maintenance binary, so the skill emphasizes safe, terminal-first commands the agent can suggest step by step while the human approves destructive actions.1.3kinstalls40Vinext Vite Nextjsvinext-vite-nextjs guides solo builders through moving a Next.js codebase onto Vite while preserving familiar APIs—file-based routing, server rendering, React Server Components, and `next/*` import patterns—so you are not locked to a single host or compiler. The skill centers on automated setup with `npx vinext init`, which runs a compatibility check, installs Vite and RSC-related plugins, adjusts module type and legacy CJS configs, and adds parallel dev/build scripts without deleting your existing Next workflow. That matters when you want edge deployment (especially Cloudflare Workers), faster local dev, or freedom from vendor-specific build pipelines but cannot afford a full framework rewrite. Use it when triggers like “migrate this project to vinext,” “convert Next.js to Vite,” or “deploy my Next.js app to Cloudflare Workers” match your session. Intermediate complexity: you should already ship a Next app and read migration warnings from `vinext check`.1.2kinstalls41Mirofish Offline SimulationMiroFish-Offline is an agent skill for solo builders who want swarm-intelligence forecasting on their own hardware. Feed a press release, policy draft, or financial report and the workflow builds a Neo4j knowledge graph, spawns hundreds of persona agents, and simulates debate and sentiment evolution over simulated social feeds—all via Ollama without paid inference APIs. Prism catalogs it for developers validating positioning before launch, prototyping PR crisis tools, or adding offline what-if labs to research products. Expect intermediate complexity: you install Neo4j Community Edition, pull Ollama models, and follow the skill’s configuration steps. It complements cloud-based opinion APIs when privacy, cost, or air-gapped demos matter. Invoke when triggers mention mirofish offline, local multi-agent simulation, or Neo4j plus Ollama agent setups. Outputs are simulation traces and narrative reports suitable for human review, not verified market predictions.1.2kinstalls42Nanochat Llm Trainingnanochat-llm-training is an agent skill that walks solo and indie builders through Karpathy’s nanochat: clone the repo, sync dependencies with uv, configure depth-driven hyperparameters, and run pretraining through finetuning, evaluation, inference, and a simple chat UI on one GPU node. It targets hackers who want reproducible, hackable LLM training without a sprawling MLOps stack—useful when you are validating a small model before productizing RAG or agents. Expect intermediate-to-advanced comfort with shells, GPUs, and long-running jobs. The skill emphasizes cost-transparent speedruns and leaderboard-style iteration rather than managed cloud training platforms.1.2kinstalls43Corridorkey Green Screencorridorkey-green-screen teaches agents to run CorridorKey: an AI keyer that separates true straight foreground color and a linear alpha from green-screen plates, including tricky semi-transparent pixels. Solo creators and small VFX shops use it when After Effects-style keys break on hair or defocus and they need EXR-friendly output. The skill covers prerequisites (uv, NVIDIA or MLX hardware), the two-input-per-frame contract (RGB image plus rough hint mask), and inference-oriented usage aligned with CorridorKey’s documentation. Complexity is advanced because gamma, gamut, and bit depth matter for clean comps. It is a specialized media integration, not a general background-removal shortcut for social clips unless you accept EXR workflow overhead.1.2kinstalls44Openviking Context DatabaseOpenViking Context Database is an expert integration skill for Volcengine’s open-source OpenViking store, which treats agent context like a filesystem rather than a fragmented vector pile. Solo builders adding durable memory, bundled resources, and skill artifacts to Claude Code–class agents use it when setup spans Python packaging, optional CLI tooling, and provider configuration. The skill walks installation (pip and cargo paths), prerequisites, and trigger scenarios such as querying the database and folding RAG into a project. It targets builders in the agent-tooling lane who need self-evolving session memory and clearer retrieval traces than a bare embedding index. Intermediate complexity reflects multi-runtime deps (Python, Go, native compile). It is not a replacement for your application UI or auth layer—it is the context layer underneath the agent.1.2kinstalls45Clui Cc Claude OverlayClui CC — Claude Overlay is an agent skill for solo builders on macOS who live in Claude Code and want a dedicated command-line UI instead of juggling terminals. It walks through prerequisites—macOS 13 or newer, Node 18+, Python 3.10+ with setuptools on 3.12+, an authenticated claude CLI, and Whisper for voice—and the install path implied by the skill (Homebrew Node, pip setuptools, npm global Claude Code). The overlay adds multi-tab sessions, a permission approval surface wired through PreToolUse HTTP hooks, voice capture, history, and a skills marketplace while staying local without telemetry. Triggers in frontmatter mirror how people search: set up Clui CC, floating overlay, permission UI, multi-tab sessions, and voice input. Use it when you are optimizing your personal agent cockpit during Build, not when you are shipping customer-facing product UI. It complements catalog skills you install into Claude—not a replacement for Prism’s directory itself.1.2kinstalls46Git City 3d Github Visualizationgit-city-3d-github-visualization is an agent skill aimed at solo builders hacking on or forking Git City, a novelty SaaS-style experience that turns GitHub stats into explorable 3D pixel buildings. The README encodes clone-and-run setup, required Supabase and GitHub credentials, and explicit trigger phrases for extending the renderer, adding achievements, or new decorations. For indie developers, it accelerates onboarding to an opinionated Next.js and React Three Fiber codebase without spelunking every module. Use it when you are actively modifying visualization logic, city features, or deployment configuration—not when you only need a static chart of GitHub metrics. Agents should respect secrets handling for GITHUB_TOKEN and Supabase service keys and test changes against the local dev server workflow documented in the skill.1.2kinstalls47Karpathy Jobs Bls VisualizerKarpathy-jobs-bls-visualizer documents how to install, configure, and extend the open-source karpathy/jobs project: scrape and parse BLS occupation pages, visualize 342 occupations in a treemap, and optionally run an LLM-powered scoring pipeline through OpenRouter to add custom layers such as AI exposure. Solo builders researching career pivots, B2B workforce tools, or education products can fork the repo, swap prompts, and regenerate color semantics without rebuilding the whole UI. The skill emphasizes uv, Playwright Chromium, and environment keys only when you need LLM scoring; base exploration can stay on built-in BLS metrics. It is a research and pipeline skill—not a hosted Prism integration—so you run it locally or on your infra. Intermediate comfort with Python tooling and API keys is expected when customizing scores.1.2kinstalls48Posterskill Academic Postersposterskill-academic-posters is a Claude Code agent skill for solo researchers and indie builders who need a conference poster without a separate design toolchain. It ingests Overleaf paper source cloned into a local overleaf directory, optionally compares against reference poster PDFs, and walks the agent through prompts for your project URL and layout specs. The outcome is a poster directory centered on index.html that is print-ready and editable in the browser. It fits the Build phase when you already have LaTeX or Overleaf content and want a fast HTML poster instead of manual PowerPoint or InDesign work. Installation is git-centric: clone the posterskill repo, clone your Overleaf project, then invoke the skill from Claude Code. Because it is procedural knowledge packaged as a command, it pairs well with agents that can run shell git operations and iterate on HTML layout with you.1.2kinstalls49Json Render Generative Uijson-render Generative UI is an agent skill for adopting the json-render framework so large language models produce dynamic interfaces as JSON rather than fragile JSX strings. Solo builders shipping agent-facing products need predictable rendering: AI chooses from a catalog you define, and json-render enforces types and component boundaries on React, Vue, Svelte, Solid, React Native, and export targets like video, PDF, and email. The skill documents installation for each stack, when to add the shadcn bridge for a large pre-built set of components, and trigger phrases for catalog creation and streamed UI. It fits early product UI where prompts change often but design systems must stay consistent. Use when integrating generative panels, copilot layouts, or multi-surface content without giving the model arbitrary code execution on the client.1.2kinstalls50Trump Code Market SignalsTrump-code-market-signals packages an open-source Trump Code workflow for solo quant-curious builders who want agent-guided access to Truth Social and X posting patterns correlated with S&P 500 moves. After cloning the repo and installing Python requirements, you use trump_code_cli.py for today’s signals, rule health, and brute-force model search outputs grounded in millions of tested combinations. The skill documents required GEMINI_KEYS for AI briefing and optional Anthropic and Polymarket keys for deeper analysis and prediction-market context. It is niche finance research—not general SEO or app growth—and fits Prism’s Idea phase when you are exploring signals, not when you need audited brokerage execution. Treat published hit rates as project claims to verify yourself; combine with your own risk controls before any live trading.1.2kinstalls51Openclaw Rl TrainingOpenClaw-RL Training is an agent skill for solo and indie builders who want reinforcement learning from natural conversation feedback instead of hand-labeled datasets. It documents a fully asynchronous framework: serve rollouts through an OpenClaw-wrapped model, capture multi-turn chats as trajectories, score turns with a PRM or judge, and train the policy in the background with GRPO or on-policy distillation while users keep chatting. The skill fits builders already running or planning a self-hosted agent API and ready to wire slime or Tinker, local GPU, and OpenClaw endpoints. Use it when you need personalized policies that improve from thumbs-up, corrections, and downstream task success—not when you only need a static prompt. Expect integration work across serving, data collection, and training jobs; this is advanced agent infrastructure, not a one-shot codegen shortcut.1.2kinstalls52Toon Formattoon-format is a journey-wide agent skill from the Daily 2026 Skills collection that teaches Token-Oriented Object Notation (TOON)—a compact, human-readable JSON data model encoding built for LLM input. Solo builders hit context limits when they paste large JSON blobs into Claude Code, Cursor, or Codex; TOON merges YAML-like indentation with CSV-style tables for uniform arrays so prompts shrink by roughly forty percent while staying legible to models. The skill documents npm/pnpm/yarn install, a global CLI for encode and decode with stdin pipes, pretty JSON output, and --stats comparisons, plus the core encode/stringify API for scripts that feed agents. Use it whenever you serialize catalogs, tool results, or config snapshots into a prompt, whether you are ideating on data shapes, validating a prototype payload, building integrations, or tightening production agent runs. It is procedural knowledge, not a hosted service: you still own validation, secrets redaction, and choosing when lossless round-trip decode is required before trusting transformed data.1.2kinstalls53Memory Lancedb Pro Openclawmemory-lancedb-pro-openclaw is an expert configuration skill for the production-grade memory-lancedb-pro plugin on OpenClaw agents. Solo builders install it when preferences, decisions, and project context must persist across sessions in a local LanceDB database instead of evaporating from the chat window. The skill covers installation (including one-click setup scripts), enabling hybrid BM25 and vector retrieval, cross-encoder reranking, multi-scope isolation, and operating memory_recall and memory_store tools. It also addresses migration, upgrades, and tuning smart auto-capture so agents surface the right memories before responding. Complexity is advanced because you are wiring embeddings, decay, and scope rules into a live agent runtime. Use during Build agent-tooling setup and Operate infra when you iterate retention and isolation in production.1.2kinstalls54Zeroboot Vm SandboxZeroboot VM Sandbox teaches agents and solo builders how to execute code inside sub-millisecond forked KVM virtual machines powered by Firecracker and copy-on-write memory mapping. Each run is a real VM with hardware memory isolation, aimed at AI workflows that must not trust generated scripts on the developer laptop or CI runner. The skill covers installation for Python and TypeScript SDKs, API key authentication, and the template-snapshot-to-fork pipeline so you can spin sandboxes when validating agent tools, running eval harnesses, or prototyping integrations. It sits primarily on the Ship security shelf because isolation is the gate before broader automation, but the same pattern applies during Build when wiring agent-tooling and during Operate when rehearsing incident reproduction in a disposable VM.1.2kinstalls55Taiwan Md Knowledge BaseTaiwan MD Knowledge Base is an agent skill for contributors to frank890417/taiwan-md, an AI-native open knowledge base about Taiwan built on Astro v5. It walks solo builders through clone, install, dev server, build, preview, and the sync step that copies `knowledge/` into `src/content/`. Triggers cover adding articles, categories, bilingual pairs, graph nodes, and local setup. The architecture treats Markdown in `knowledge/` as the single source of truth while the public site and D3 knowledge graph are generated outputs. Use it when extending structured cultural or reference content—not when building unrelated SaaS products. Prerequisites are Node.js 18+ and npm or pnpm.1.2kinstalls56Slug Font RenderingSlug Font Rendering is an agent skill that packages procedural knowledge for the Slug algorithm: high-quality GPU text by resolving glyph outline coverage in fragment shaders. Solo builders shipping games, creative tools, or custom renderers use it when bitmap atlases blur at zoom, bust VRAM, or fail CJK scale requirements. The skill points to authoritative sources—the JCGT paper and Terathon blog updates—and describes what Slug does: encode outlines as Bézier curves and segments, compute coverage per pixel, and render through any API that supports the supplied shader model. Triggers align with implement-Slug-on-GPU tasks, HLSL porting, and replacing texture-based text pipelines. It is reference-heavy rather than a one-click generator, so you still wire draw calls, font loading, and buffer layouts for your engine or framework.1.2kinstalls57Grimmory Self Hosted LibraryGrimmory Self-Hosted Library is an agent skill that walks solo builders through installing, configuring, and extending Grimmory—the BookLore successor—for a private book collection. It assumes Docker and Docker Compose, documents required environment variables (app IDs, MariaDB credentials, timezone, disk type), and explains how to wire persistent storage and database services so the app stays reliable on a home server or VPS. Beyond bare install, the skill covers operational features builders care about when self-hosting media: OPDS for third-party readers, Kobo and KOReader sync, metadata lookup, bookdrop imports, and multi-user access without renting a commercial locker. It is aimed at indie operators who want full control over ebooks and comics rather than shipping a SaaS product. Use it when triggers like “set up Grimmory,” “grimmory docker setup,” or “self-hosted book library” match your session, especially before you point a domain or reverse proxy at the stack.1.2kinstalls58Translate Book ParallelTranslate Book (Parallel Subagents) is a Claude Code skill for solo builders and authors who need entire books converted to another language without single-session context collapse. It ingests PDF, DOCX, or EPUB, normalizes through Calibre and Markdown, splits text into roughly six-thousand-character chunks tracked in a manifest with SHA-256 hashes, and fans work to parallel subagents that each read one chunk and write translated markdown. After hash validation and a strict one-to-one chunk match, it merges via Pandoc into HTML with table of contents and exports DOCX, EPUB, or PDF again through Calibre. The design targets truncation and context accumulation failures common in monolithic translation chats. You need Calibre and Pandoc installed locally; the skill is heavy on filesystem and shell orchestration. It suits licensing-compliant personal or client manuscripts, not ad-hoc paragraph paste, and assumes you will review linguistic quality after the mechanical pipeline completes.1.2kinstalls59Witr Process Inspectorwitr Process Inspector teaches your coding agent to use witr—Why Is This Running—a Go CLI and TUI that reconstructs how a process or listening port came to exist. Instead of manually chaining ps, lsof, ss, systemctl, and docker ps, you invoke witr with natural triggers when a port is occupied or a daemon will not die. The skill documents installation on Unix and Windows, package managers, and when to prefer CLI versus TUI for tracing supervisor and container chains. Solo builders benefit during Operate when staging and production diverge, and during Ship when debugging integration tests that leave servers bound. It is a thin integration skill: the heavy lifting is the external witr binary, while the agent learns invocation patterns and interpretation of causality output.1.2kinstalls60Kimodo Motion DiffusionKimodo Motion Diffusion is an agent skill that wires your coding agent to Kimodo, NVIDIA’s kinematic motion diffusion stack for generating 3D human and humanoid robot movement from natural-language prompts plus explicit kinematic constraints. Solo and indie builders who touch robotics sims, embodied agents, or character-heavy games can use it to stand up walking, manipulation, or path-following motions quickly instead of blocking on capture gear or hand-keyed cycles. The skill documents clone-and-pip or Docker install, GPU expectations, and trigger phrases for interactive demos, MuJoCo-oriented G1 output, and constrained end-effector control. It fits the validate phase when you need a motion clip or trajectory artifact to test narrative, UX, or sim fidelity before you harden backend pipelines or ship polished media. It is niche relative to typical SaaS stacks but high leverage when motion quality gates your prototype.1.2kinstalls61Autoresearch Genealogyautoresearch-genealogy is a structured research kit for solo builders who use Claude Code to investigate family history without ad-hoc chat threads. It packages autonomous autoresearch prompts, an Obsidian vault scaffold, regional archive pointers, and methodology guardrails so your agent can iterate on sources, write citations into notes, and audit contradictions. The audience is not typical SaaS shipping—it is builders treating genealogy like a long-horizon knowledge product where rigor matters as much as speed. Install it when you want repeatable loops for ancestor discovery, document organization, and DNA-informed analysis rather than one-off summaries. It also adapts to manual workflows if you only borrow the templates. On Prism it sits in Idea research but spans documentation and quality checks that resemble Validate and Operate-style auditing for your personal vault.1.2kinstalls62Nightingale KaraokeNightingale Karaoke is an agent skill for builders who want a local, ML-powered karaoke experience without stitching Demucs, Whisper, and a game UI themselves. It documents Nightingale—a Rust Bevy app that scans your music folder, separates vocals from instrumentals, transcribes lyrics with timestamps, and plays back with highlighting, pitch scoring, and optional shader backgrounds. The skill walks through pre-built downloads from GitHub releases or compiling from source, including Linux ALSA/Wayland dependencies and macOS quarantine removal. It explains the automatic first-run bootstrap that pulls ffmpeg, Python, PyTorch, and models so solo operators are not manually conda-installing on day one. Use it when you are building or operating a personal karaoke library, debugging Nightingale profiles, or evaluating stem-separation quality—not for generic Rust game tutorials. Triggers in the skill readme map natural phrases like adding karaoke to a music library or troubleshooting setup. Expect shell commands, large model downloads, and GPU considerations.1.2kinstalls63Modly Image To 3dModly Image-to-3D is a build-focused skill for solo makers who need printable or game-ready meshes from photos without renting a hosted 3D API. It explains Modly’s architecture—an Electron shell talking to a local FastAPI service that runs Hunyuan3D-style generators on the user’s GPU—and how extensions ship as manifest plus Python generator modules. Indie game devs, ecommerce mockup builders, and technical artists use it when privacy, offline runs, or GPU cost control matter more than fastest cloud latency. The readme covers installation paths from launcher scripts through dependency setup implied by api/requirements.txt, and how frontend IPC bridges the renderer to generation jobs. It does not replace art direction or mesh cleanup in a DCC tool; it gets a local pipeline running so agents and developers can iterate on image-to-3D inside a desktop repo structure they can fork and extend.1.2kinstalls64Keyid Agent Kit McpKeyID Agent Kit MCP is a build-phase integration skill for solo builders who want their coding agents to send and receive real email instead of simulating inbox actions in chat. It documents how to install @keyid/agent-kit, run the stdio MCP server, and configure Ed25519 keys so Claude Desktop, Cursor, or any MCP-compatible client gains 27 structured tools for messaging workflows. The skill fits indie operators automating support triage, founder outreach, or agent-driven CRM touches where a durable email identity matters. Unlike a generic SMTP snippet, it standardizes MCP tool names and server startup so agents can reply, forward, search threads, and manage contacts with predictable contracts. You should pair it with your own security review because live email access implies network reach and sensitive message content; Prism does not substitute for reading audits on this listing before enabling production inboxes.1.1kinstalls65Fabro Workflow FactoryFabro Workflow Factory teaches agents how to adopt Fabro, an open-source AI coding workflow orchestrator in Rust. Solo builders and tiny teams use it when they outgrow ad-hoc “run Claude then run Codex” chats and need durable pipelines: Graphviz DOT graphs that encode branching, loops, human approval gates, multi-model routing, and cloud sandbox runs behind a persistent service. The skill covers installation (curl installers aimed at Claude Code, Codex, or Bash), global fabro install, per-project fabro init creating .fabro/ config, and the CLI surface for authoring and running workflows. Triggers align with setting up Fabro, writing DOT graphs, adding human-in-the-loop gates, and configuring routing. It fits Prism’s Build journey for agent-tooling and extends into Operate when the same graphs become production runbooks. Intermediate complexity reflects DOT literacy and service operations. After init, your deliverable is a runnable graph and project config, not a single chat transcript.1.1kinstalls66Aracli Deploy ManagementAraCLI Deploy Management is a practical agent skill for solo builders and small teams running OpenClaw-compatible AI agents who need a sober infrastructure map before paying for the wrong hosting tier. It walks through cloud VMs with familiar Terraform and Ansible patterns, lighter managed container platforms, and bare-metal or hybrid options where latency and GPU economics matter. The guide surfaces concrete starter commands such as docker compose for agent runtime, then frames management modes—terminal CLI for scripts, APIs for automation, MCP for agent-to-agent control—with honest pros and cons on scaling, pricing, and operational burden. Use it when moving from a laptop prototype to production, comparing Hetzner-style VMs against Fly or Railway, or deciding how operators will restart, observe, and patch live agent fleets.1.1kinstalls67Openhanako Personal Ai AgentOpenHanako Personal AI Agent is a procedural skill for solo builders who want a local, always-on assistant instead of only chat-in-the-IDE. It walks through downloading or building the OpenHanako Electron app, then configuring agents with durable memory, personality prompts, and tool access across the filesystem, shell, browser, and scripted automation. The skill emphasizes practical extension: adding skills to Hanako, wiring Telegram, scheduling recurring work with cron, and running multiple agents that coordinate in shared channels. That makes it useful when you are productizing a personal copilot, prototyping multi-agent workflows, or dogfooding agent UX before shipping a SaaS agent feature. Complexity is intermediate because you need comfort with npm builds, OS install quirks (Gatekeeper, SmartScreen), and safe delegation of machine access. It is not a hosted API integration skill—it is the playbook for owning the full desktop agent stack.1.1kinstalls68Codex Autoresearch LoopCodex Autoresearch Loop is an agent skill that runs Karpathy-style autoresearch on everyday software goals: you state a measurable target in one sentence, Codex confirms the approach, then iterates without you—changing code, running verification, keeping improvements in git, and reverting bad steps. Solo and indie builders use it when they want overnight progress on flaky tests, type errors, coverage floors, or other objective metrics without babysitting the agent. It fits the build phase when the product already exists and the bottleneck is relentless iteration, and it bleeds into ship when verification is tests or static checks. The workflow is explicitly Codex-oriented (daily 2026 skills collection) but the pattern—goal, verify gate, git discipline—maps to any repo-backed agent setup. Best when the goal is objectively checkable; weak when success is subjective design judgment.1.1kinstalls69Flash Moe Inferenceflash-moe-inference documents the Flash-MoE engine for solo builders who want frontier-scale Mixture-of-Experts models on a MacBook instead of only cloud GPUs. The skill walks through hardware requirements (Apple Silicon, 48GB+ RAM, ~210GB free SSD), cloning danveloper/flash-moe, building via Makefile, and preparing weights from HuggingFace safetensors with Python used only for extraction. It targets agents when users ask to run huge LLMs locally, stream experts from NVMe, or use Metal for quantized MoE on macOS. Advanced complexity reflects OS version gates (macOS 26+ / Darwin 25+), large downloads, and Metal toolchain needs. After you have a working chat binary, Operate-phase monitoring is mostly your own process management—this catalog entry sits primarily on the Build shelf as specialized agent tooling for local inference R&D.1kinstalls70Ghostling Libghostty TerminalGhostling — libghostty Terminal Emulator is an agent skill for developers who want a minimal, embeddable terminal without forking a full terminal app. It explains how Ghostty’s libghostty-vt C API handles SIMD-optimized VT sequence parsing, cursor and style state, scrollback, Unicode graphemes, and render-state diffs, while you supply windowing and drawing—here via Raylib in a compact C example. The skill is triggered when you need to integrate ghostty’s library, build a small emulator, or understand what the API does and deliberately does not cover. Solo builders creating dev tools, custom IDEs, game debug consoles, or agent-facing local UIs can use it as a procedural map from clone-through-build to hooking keyboard and mouse protocols. Expect advanced systems work: CMake and Ninja builds, compiler setup, and clear separation between terminal core and your PTY and platform layer. It is not a finished product template for tabs and settings; it is the integration blueprint for owning the shell around libghostty-vt.1kinstalls71Code Review GraphCode-review-graph builds and maintains a structural knowledge graph of your repository using Tree-sitter, stores it in SQLite, and serves it through an MCP server so Claude queries relationships and reads only files in the blast radius of a change. Solo builders drowning in monorepo context costs get a practical alternative to dumping entire trees into every task. Setup paths include a Claude Code marketplace plugin or pip plus MCP registration; Python 3.10+ and uv are expected. Day-to-day use spans pre-review graph builds, incremental updates after commits, and dependency-aware questions about what your edit might break. It is an integration skill, not a linter: it changes what gets loaded, not whether code passes rules. Large repos and frequent reviews benefit most; tiny codebases may not justify graph maintenance.977installs72724 Office Ai Agent7/24 Office AI Agent is an agent skill for solo builders who want a framework-free, always-on assistant in Python. It walks through cloning the 724-office repo, installing minimal dependencies, laying out workspace/memory directories, and filling config.json with default and embedding model endpoints. The system advertises 26 tools, three-layer memory, MCP server attachment, cron-based recurring tasks, runtime-defined tools, and self-repair flows—useful when you outgrow ad-hoc scripts but do not want a heavy agent framework. Triggers cover MCP wiring, memory configuration, WeChat Work integration, and production scheduling. Expect to operate the same stack in Grow or Operate phases once cron and diagnostics keep the agent healthy overnight.974installs73My Brain Is Full CrewMy Brain Is Full Crew is an install skill that drops ten coordinated Claude Code agents into your Obsidian vault’s `.claude/` directory so solo builders and researchers can manage notes, nutrition, and mental wellness through conversation instead of manual filing. Triggers cover setup phrases like installing the brain crew, configuring Obsidian agents, or initializing a second brain. Architect handles structure; Scribe cleans captures; Sorter nightly triage; Seeker searches and synthesizes; Connector surfaces hidden links; Librarian runs weekly vault hygiene. The skill is phase-specific agent tooling: you use it when standing up or reconfiguring personal knowledge infrastructure, then live with the crew during daily Operate-style note work without redeploying an app. It complements generic coding skills by targeting life/knowledge ops for people drowning in inbox notes. Prism lists it for builders who already live in Obsidian and want agent specialization without writing ten prompts from scratch.960installs74Claude Peers Mcpclaude-peers-mcp packages an MCP server and setup ritual so solo builders running multiple Claude Code instances on the same machine can discover each other and swap messages in real time. A lightweight broker daemon listens on localhost:7899, persists peer state in SQLite, and routes HTTP traffic while each session’s MCP server forwards inbound messages into the active Claude channel. The skill walks through cloning the repo, installing dependencies with bun, registering a global stdio MCP entry, and launching Claude Code with the development channel that exposes the peers server. It fits indie builders who split work across parallel agent sessions—research in one window, implementation in another—and need coordination without copying chat by hand. It is not a hosted team chat product; everything stays on your machine. Expect elevated permission flags during setup because the channel loader is explicitly marked dangerous for development use.954installs75Weixin Agent Sdkweixin-agent-sdk is a TypeScript framework that bridges any AI agent backend to WeChat (微信) through the Clawbot channel. Solo builders shipping bots or support agents for Chinese users install it when they need a working message loop without standing up HTTPS webhooks on a public URL. The skill walks through npm or pnpm installation, a one-time QR login that persists credentials locally, and implementing the small Agent interface so your OpenAI, Claude, or custom logic handles inbound text and returns replies. Long-polling receives messages, which keeps deployment simple for indies who run workers on a laptop, VPS, or background process. It fits the Build phase as a concrete integration: authenticate once, implement chat, call start, and iterate on prompts and tooling around the same loop. Expect to manage Node 22+, network access for WeChat, and secure handling of stored login state on the machine that runs the bot.934installs76Polymarket Arbitrage Trading BotPolymarket Arbitrage Trading Bot is a skill package for builders automating prediction-market edges on Polymarket’s short-interval crypto markets. It describes cloning the apechurch TypeScript repo, installing dependencies, copying .env.example, and running npm build—with Node 16+, Polygon USDC, and a Polymarket-compatible wallet for live mode. The strategy polls prices, detects dumps on one leg, and buys both sides when combined cost sits under a configured threshold to lock structural edge ahead of resolution. Triggers cover setup, configuration, dump-hedge runs, simulation, and CLOB arbitrage phrasing. Treat this as integration-heavy backend automation with real funds risk: validate simulation, keys, and compliance yourself before production trading.863installs77Dingtalk Workspace CliDingTalk Workspace CLI (dws) is an agent skill for solo and indie builders who need DingTalk in scripts and coding agents without clicking through the desktop app. DingTalk open-sourced dws as a single binary that wraps the workspace product suite; every command returns structured JSON so you can pipe results into automation or let an LLM parse outcomes reliably. The skill documents recommended installation via curl or PowerShell installers, PATH setup, and when to reach for dws versus manual admin work—searching contacts, creating calendar events or todos, driving attendance and approvals, or sending messages as part of an AI agent workflow. It fits builders shipping internal tools, ops automations, or agent products that must talk to DingTalk corporately. Pair it with your repo’s secrets handling and deployment story; dws does not replace product UX but removes repetitive workspace chores from the build and operate loop.857installs78Web Access Claude SkillWeb Access Claude Skill packages a complete internet stack for Claude Code: when to search versus fetch versus drive a real browser, how to connect over Chrome DevTools Protocol, and how to fan out parallel sub-agents for heavy research. Solo builders hit the gap where built-in WebSearch and WebFetch lack click-through flows, authenticated sessions, and remembered site quirks—this skill documents the dispatch playbook and CDP proxy pattern to close that gap. Install it when you are building agents that must verify docs, file forms, scrape dynamic pages, or run multi-tab investigations without rewriting the same glue every sprint. It assumes you can run local Chrome and accept the security tradeoffs of browser automation. It is not a hosted scraping SaaS; it is procedural knowledge for your agent runtime. Pair it with careful secret handling because inherited browser state may include logged-in accounts.845installs79Weclaw Wechat Ai BridgeWeClaw WeChat AI Bridge is an agent skill for indie builders who want WeChat as a front door to coding agents and LLM backends without building a custom protocol stack. The Go service handles QR-code authentication, routes inbound chat to your chosen agent, converts media, and manages daemon lifecycle with straightforward CLI commands. You can run agents through ACP for low-latency JSON-RPC subprocesses, spawn CLI processes per message, or point at HTTP endpoints that speak OpenAI-compatible REST. Installation paths include a one-line installer, go install from GitHub, or a Docker image with a persisted config directory. The skill fits Validate prototypes that demo chat-driven workflows and Build integrations where support or ops happens in WeChat. It does not replace your model provider keys or agent prompts—you still configure backends and security boundaries. Outcome: a running bridge so messages from WeChat reach your AI agent with documented start, login, and status flows.841installs80Awesome Phd CvAwesome PhD CV is an agent skill that steers LaTeX resume and CV authoring for researchers and PhD students who need the right format for a specific gate—faculty packets, postdoc listings, or ATS-scanned industry applications. Solo and indie builders who also wear a researcher hat use it to pick among Jake's single-column ATS layout, Deedy's dense two-column profile, or Awesome-CV's multi-page academic record instead of reinventing class files from scratch. The skill maps each template to engine choice, column layout, and page budget so compilation errors and font issues are caught early. It aligns with triggers such as LaTeX PhD CV help, Deedy or Jake's templates, and converting an academic CV to a big-tech-safe resume. Install when you are actively job hunting or refreshing credentials; skip when you only need a plain Markdown bio for a landing page.839installs81Awesome Free Llm ApisAwesome Free LLM APIs is a reference agent skill that catalogs LLM vendors offering lasting free tiers for text inference, with rate limits, flagship models, and region constraints in comparable tables. Solo builders use it when bootstrapping agents, chat features, or CI smoke tests without burning paid credits or chasing expiring trials. Coverage spans provider-native APIs such as Cohere, Google Gemini, and Mistral alongside inference hosts like Cerebras that serve open-weight stacks. The skill emphasizes OpenAI-compatible integration patterns so you can point familiar SDKs at alternate base URLs after you pick a quota that fits your traffic. It fits early Validate spikes and Build integration work; it is not a substitute for production SLAs, compliance review, or cost modeling at scale. Pair it with your own secret handling and retry logic once you exceed documented RPM or monthly caps.833installs82Filmkit Fujifilm CameraFilmKit Fujifilm Camera is a Prism skill for builders and creators who work with Fujifilm X-series bodies and want agent help around a zero-install browser tool rather than desktop-only vendor apps. FilmKit uses WebUSB and the PTP protocol—the same family as Fujifilm X RAW STUDIO—so RAW conversion stays on the camera processor while the page manages presets and previews. The skill explains preset slots, import/export, RAF auto-detection, and mobile usage patterns so an agent can troubleshoot connections, preset workflows, or feature limits. It suits indie makers blending photography products, creative tooling, or hardware-adjacent web experiments, not typical SaaS backend work. Invoke when triggers mention FilmKit, Fuji profiles, WebUSB, or in-browser RAF conversion.826installs83Vercel Labs Emulatevercel-labs-emulate is an agent skill for solo builders who integrate with Vercel, GitHub, or Google but cannot hit real APIs in CI or locked-down environments. It documents emulate, a drop-in local HTTP stack that keeps real semantics—state, OAuth flows, webhooks, and destructive cascades—so Vitest and sandbox runs fail for the right reasons. Install via npx or devDependency, start all or selected services, seed from YAML, and point your SDK base URLs at localhost ports. Use it when triggers like emulate vercel api locally, mock github api for tests, or no-network sandbox testing match your pipeline. It reduces flakiness compared to hand-rolled mocks and avoids shipping code that only passed against stubbed responses. The skill is procedural knowledge for wiring tests and local dev, not a hosted emulator product; pair it with your existing deploy and integration skills once tests green-light behavior.819installs84Opencli Rs Web ScraperOpencli-rs-web-scraper documents how solo builders install and use opencli-rs, a compact Rust CLI that pulls real-time information from dozens of global platforms with one command pattern. It fits agents and terminal-first workflows: fetch Hacker News tops, Bilibili trending, or social threads for validation research, then reuse the same tool during Build to integrate scraping and desktop automation into scripts or agent skills. Session reuse via a Chrome extension reduces auth friction compared to brittle cookie copying. The skill also covers generating YAML adapters, AI-assisted API discovery, and controlling Electron apps like Cursor from the shell, plus delegating to familiar CLIs such as GitHub CLI, Docker, and kubectl. Triggers in the skill mirror install paths for macOS, Linux, and Windows and everyday commands indie hackers actually type. Treat network fetching and desktop control as privileged operations: scope what sites and apps you automate and keep credentials inside your browser profile. This is an integration skill package, not a hosted scraper service.817installs85Creationix Rx Data Storecreationix-rx-data-store teaches coding agents how to use RX (@creationix/rx), an embedded store for JSON-shaped data that keeps documents queryable after a single encode. Solo builders shipping Node or TypeScript backends, CLIs, or agents that repeatedly open the same large JSON blobs—deployment manifests, catalogs, config snapshots—can avoid re-parsing entire trees on every read. The skill covers installation, the core stringify/parse string API, and the mental model: encode once, then perform random-access reads with minimal heap pressure, positioned as no-SQL SQLite for unstructured payloads. It is for implementation time when you are choosing serialization and read patterns, not for product discovery or marketing. Use it when triggers match encoding, parsing, or querying RX buffers; skip when a relational database or tiny in-memory JSON is enough.813installs86Cc Connect Ai BridgeCC-Connect AI Bridge teaches solo builders how to pipe locally running coding agents into the messaging apps they already live in. Instead of staying glued to a terminal, you configure cc-connect to relay prompts and replies through Telegram, Slack, Discord, Feishu, DingTalk, WeChat Work, LINE, or QQ. The skill covers stable and beta npm installs plus curl-based binaries for Linux amd64 and macOS arm64, matching typical indie dev machines. It fits anyone running Claude Code, Codex, Cursor Agent, Gemini CLI, or sibling CLIs who wants remote triage, async reviews, or lightweight task scheduling from mobile chat. Setup is integration-focused: install the bridge, bind your agent, and tune per-platform bot configuration. Expect operational follow-through—tokens, webhooks, and local agent health remain your responsibility after the bridge is up.807installs87Microsoft Rust TrainingMicrosoft Rust Training is an agent skill that catalogs Microsoft’s structured Rust curriculum for solo builders who need a ordered path from another language into production-grade Rust. Three bridge books orient C/C++, C#/Java/Swift, and Python developers toward ownership, RAII, static typing, and GIL-free concurrency; a dedicated async book centers Tokio, streams, and cancellation safety; advanced pattern and type-driven correctness books cover Pin, unsafe, allocators, phantom types, and capability tokens; and an engineering practices book closes the loop on how teams ship maintainable Rust. Each volume spans roughly fifteen chapters with diagrams, editable playgrounds, and exercises so agents can assign reading plus hands-on tasks instead of hallucinating lesson order. Invoke it when onboarding to Rust for a CLI, API, or systems side project, or when an agent must pick the right book for a background (e.g. Python → static typing) before writing code.786installs88Jkvideo Bilibili React Nativejkvideo-bilibili-react-native is an agent skill for solo builders who want a production-shaped Bilibili-like experience in React Native rather than a minimal WebView wrapper. It documents JKVideo’s architecture: DASH adaptive video, danmaku overlays, WBI-signed REST calls, QR authentication, live rooms with WebSocket danmaku, and offline downloads with LAN sharing. Triggers match real implementation tasks—adding quality ladders, fixing signature headers, or standing up Expo dev builds when Expo Go cannot load native playback features. The skill assumes you will clone the upstream JKVideo repo, install dependencies, and choose Expo Go for fast iteration or a dev build when you need full codec and resolution support. It suits indie clients, regional forks, or learning projects, not official Bilibili partnerships. Pair it with your own compliance review because third-party API usage and platform ToS apply outside this skill’s scope.785installs89Tg Ws Proxy Telegram Socks5TG WS Proxy is a setup skill for builders who rely on Telegram Desktop but hit ISP or regional blocking on raw MTProto. The agent walks through cloning Flowseal’s tg-ws-proxy, installing with pip, and pointing Telegram at 127.0.0.1:1080 while the proxy upgrades eligible flows to TLS WebSockets toward the right data center. No third-party VPN server is required—the tunnel runs on your machine. Use the console entry for headless setups or platform tray commands when you want a persistent local service. This is operational plumbing: it keeps support channels and community DMs usable while you ship, not a step in your app’s CI pipeline.784installs90Ai Scanner Garakai-scanner-garak is an agent skill for standing up 0din-ai’s open-source AI Scanner: a Ruby on Rails application that orchestrates NVIDIA garak behind a multi-tenant web UI. Solo builders shipping chatbots, agents, or custom model endpoints can install via curl-based Docker scripts, configure secrets and Postgres in `.env`, and run broad safety sweeps without hand-rolling garak CLI recipes. The skill documents minimum environment variables, installation paths, and invoke triggers for vulnerability scans, OWASP LLM Top 10 checks, scheduled assessments, and downstream reporting. It suits indie teams who need repeatable LLM red-team coverage before launch and in Operate, but who want operational guardrails instead of ad-hoc prompt hacking in a notebook.782installs91Agent Flow VisualizationAgent Flow Visualization is a VS Code extension skill that turns Claude Code’s opaque multi-step runs into an interactive node graph. Solo builders install Agent Flow from the Marketplace, open it from the command palette, and stream execution while tool calls branch, subagents coordinate, and timings accumulate—useful when a single prompt fans out into dozens of hidden steps. Frontmatter triggers cover debugging tool chains, monitoring sessions, and configuring Claude Code hooks for visualization. Requirements are VS Code 1.85+, Claude Code CLI on PATH, and Node.js for Claude Code itself. It does not replace logging infra for production; it accelerates local comprehension of orchestration bugs, wrong tool order, and runaway subagent trees. After you stabilize flows, move to Ship testing patterns or dedicated observability for deployed agents.780installs92Wewrite Wechat Ai Publishingwewrite-wechat-ai-publishing is a workflow skill for solo builders and creators targeting WeChat Official Accounts who want agent-driven articles without juggling separate research, writing, SEO, image, and upload tools. It orchestrates hotspot sourcing, topic selection, long-form writing, keyword-oriented SEO passes, cover and inline image generation via Doubao or OpenAI, WeChat-specific formatting, and submission to the platform draft box. Install it as a Claude Code skill under ~/.claude/skills, configure WeChat appid/secret and API keys, then invoke with natural Chinese commands or explicit wewrite triggers. It fits growth and launch motions where organic Chinese social distribution matters, but it assumes you operate a configured Official Account and accept third-party API usage. Components can run standalone outside the agent for automation hooks. Review secrets handling and provider terms before production publishing.771installs93Holyclaude Ai WorkstationHolyClaude AI Workstation is an agent skill for solo builders who want a reproducible AI coding lab in Docker instead of stitching CLIs, browsers, and UIs on the host. The skill walks through prerequisites, a minimal docker-compose service definition (ports, shm_size, PUID/PGID), volume layout, and operational triggers from the ara.so Daily 2026 collection. It targets developers shipping with Claude Code and peer agents who need a consistent remote or local workstation—especially when local installs conflict or you want one image to share across machines. Use it when you are standing up or debugging the containerized stack, not when you are choosing product features or marketing pages. Complexity is intermediate because you must understand Docker networking, credentials, and volume persistence.770installs94Mcp Brasil Public Apismcp-brasil-public-apis documents the mcp-brasil Model Context Protocol server so solo builders can give Claude, GPT, Copilot, Cursor, and similar agents governed access to Brazilian public datasets. Instead of hand-rolling dozens of REST clients, you install mcp-brasil, register the server in your desktop or IDE MCP config, and let the agent call hundreds of structured tools for economy, legislation, transparency, judiciary, elections, environment, and health. The skill lists trigger phrases, installation via pip or uv, and Claude Desktop configuration using uvx to run mcp_brasil.server. Most endpoints are keyless; a small subset expects free API keys in environment variables. It fits indie agents, internal copilots, and research bots aimed at Brazil-specific compliance, journalism, or fintech workflows where authoritative government sources must be queryable in chat.763installs95Type4me Macos Voice InputType4Me macOS Voice Input is an ara.so Daily 2026 skill that documents how to build, configure, and extend a Swift-based macOS dictation tool. Solo builders shipping voice-driven agent workflows use it when they need global hotkey transcription with a choice of on-device Sherpa streaming/offline clients or streaming cloud ASR, plus optional LLM cleanup before text lands in the focused field. The skill maps the repo layout—ASR abstraction, per-provider configs, Sherpa C bridge, and registry patterns—so you can add providers, wire Volcengine or Deepgram, troubleshoot recognition failures, and compile from source without guessing folder responsibilities. It matters for indie macOS developers who want Copilot-style voice input without sending audio history to a vendor by default, while still allowing cloud engines when latency or accuracy demands it.762installs96Obliteratus AbliterationOBLITERATUS is an open-source abliteration toolkit for solo builders and researchers who need to study or alter refusal behavior in transformer LLMs using mechanistic interpretability. The skill walks agents through installing obliteratus (core, spaces, or full extras), satisfying Python 3.10+, PyTorch, transformers, accelerate, and Gradio requirements, and using HuggingFace credentials when models are gated. It explains how the toolkit finds refusal directions in hidden states, surgically removes them from weights, and aims to preserve general language skills—then surfaces UI, CLI, API, and notebook entry points. Use it when trigger phrases mention abliterating a model, removing refusal, uncensoring with OBLITERATUS, or analyzing refusal geometry—not for everyday app shipping. The audience is advanced builders experimenting with local or custom models, red-team research, or interpretability—not teams looking for hosted safety-compliant APIs.760installs97Flipoff Split Flap DisplayFlipOff Split-Flap Display Emulator is an agent skill for solo builders who want a free, offline-capable retro departure-board effect in the browser. It walks through cloning the FlipOff repo, serving index.html (recommended for audio), and tailoring the flip grid, copy, colors, sounds, and keyboard interactions for either an on-page embed or a dedicated TV display. The underlying app is framework-free—no React, no npm install—so it fits indie landing pages, events, studios, and side projects where you want distinctive motion UI without shipping a heavy frontend stack. Triggers match practical tasks: add a split-flap module to a site, change quotes, resize the board, or turn a display into a flip board. Review SKILL.md and the upstream repo before exposing user-generated message content in production.756installs98Mac Code Local Ai Agentmac-code-local-ai-agent walks solo builders through running a free, local AI coding agent on Apple Silicon—positioned as a Claude Code alternative that keeps prompts on-device. The skill covers prerequisites (Homebrew stack), choosing llama.cpp versus MLX, quantization choices that fit 16 GB machines, and how the agent routes each turn into search, shell, or chat paths with concrete file and command tools. It matters when subscription costs, privacy, or offline work push you off hosted agents but you still want tool use for everyday repo tasks. Expect intermediate setup: model downloads, backend tuning, and RAM-aware quantization tradeoffs. Use across the journey whenever you want a $0/month agent driver; canonical catalog placement stays under Build agent-tooling because installation and backend wiring are the first heavy lift.751installs99Markit Markdown ConverterMarkit Markdown Converter documents how to use markit-ai to turn almost any source into markdown: office documents, notebooks, feeds, archives, web pages, and rich media. Solo builders install the global CLI or add the package to a Node, Bun, or pnpm project, then run one command per asset or wire the SDK for batch pipelines. Optional LLM keys unlock descriptions and transcriptions on images and audio. The skill fits when you are consolidating research PDFs, competitor pages, or legacy Word specs into git-friendly markdown for agents and static sites. It is an integration playbook around a specific tool—not a generic pandoc cheat sheet—so prerequisites include Node and comfort storing API keys for AI features when you need them.751installs100K Skill Korean Ai Toolsk-skill-korean-ai-tools is a collection of automation skills for Korean services, packaged so solo builders can let coding agents handle SRT and KTX reservations, KBO scores, lotto draws, KakaoTalk messages, Seoul subway arrival data, HWP file conversion, and postal code lookup. It fits indie developers and Korean-speaking founders who already use Claude Code, Codex, or Cursor and want agent-native shortcuts instead of juggling separate apps. Install the full set with npx k-skill install or pick modules, then run k-skill-setup to configure sops and age for secrets rather than pasting credentials into the chat. The skill emphasizes secure environment variables and encrypted secret files. It is multi-phase in practice: you integrate during build, then invoke the same skills during operate and personal productivity workflows when tickets or messages need handling.750installs101Legalize Es Spanish LegislationLegalize ES Spanish legislation is an agent skill for working with the legalize-dev/legalize-es repository: Spanish public law stored as Markdown, versioned like code. Solo builders shipping products for Spain—or any agent task touching BOE norms—use it when triggers fire for searching Spanish law, querying BOE identifiers, finding articles, checking regulations, or comparing versions over time. After clone, workflows stay in the terminal: grep for phrases, open BOE-A-* files by identifier, and use native Git to see how a statute changed between reforms. That turns legislative archaeology into repeatable agent steps instead of manual BOE browsing. It is a research and compliance adjunct, not legal advice; pair it with human counsel for binding interpretations.734installs102Pump Analyzer SolanaPump Analyzer Solana is an agent skill for cloning and extending a static monitoring platform aimed at tokens launched on Pump.fun on Solana. Solo builders who trade or research memecoins get a ready-made front end that streams live market data over Pump.fun WebSockets, visualizes trends, fires custom alerts, and optionally connects a Solana wallet without custodial risk. The stack is deliberately dependency-free HTML, CSS, and JavaScript so you can ship a dashboard in minutes with git clone, open index.html, or any static file server. Triggers such as set up pump analyzer, track pump.fun tokens, build memecoin dashboard, and integrate pump.fun websocket map directly to implementation tasks. It fits indie builders validating launch momentum or operating a personal watchlist before committing to a heavier SaaS stack, while keeping deployment as simple as serving static files.731installs103Turboquant PytorchTurboQuant PyTorch is an agent skill for solo builders and small teams shipping LLM-powered products who hit GPU RAM limits on long contexts. It encodes a from-scratch PyTorch reading of Google’s TurboQuant: Stage 1 applies a random orthogonal rotation and Lloyd-Max scalar quantization; Stage 2 adds QJL residual correction so attention inner products stay unbiased even when stored vectors look wrong. The skill is for implementers who already run inference in PyTorch and want 3-bit as a practical default (~5x compression with ~99.5% attention fidelity in the documented framing), not for people who only need a hosted API with default caching. Triggers include compress KV cache, 3-bit compression, QJL residual correction, and LLM inference memory optimization. Expect advanced linear algebra and quantization literacy; outputs are implementation guidance and compression ratio expectations tied to named baselines in the skill text.722installs104App Store Screenshots GeneratorApp Store Screenshots Generator is an agent skill from the ara.so Daily 2026 collection for solo builders who need store-ready iPhone marketing images without hiring a designer for every release. Before writing code, the workflow asks about your app’s brand, hero features, and visual style, then scaffolds a lean Next.js project that renders each slide as a polished ad with compelling copy. Slides use proven App Store messaging patterns, device frames, and html-to-image export so you get PNGs at all four Apple-mandated resolutions. Multi-locale and RTL support help indie apps ship localized listings faster. Use it in Launch when you are preparing an App Store submission or refreshing screenshots for a growth push. It complements ASO keyword work by improving conversion on the visual funnel; it does not replace App Store Connect metadata, pricing, or review compliance checks.714installs105Open Pencil Design Editoropen-pencil-design-editor documents OpenPencil, an AI-native open-source alternative to Figma that keeps native .fig compatibility while exposing automation surfaces solo builders actually use: a global CLI, an MCP server for Claude/Cursor-style agents, and a Vue SDK for embedding editors. Install paths span zero-setup web demo, Homebrew desktop, and bun-based CLI/MCP packages from @open-pencil. For indie founders, it bridges Validate-phase lo-fi prototypes and Build-phase screen implementation without paying for seat-based design tools or brittle export pipelines. Agent users wire MCP so tasks like inspecting layers, batch edits, or scripted file updates happen headlessly. The skill is editorial coverage of ara.so Daily 2026 Skills rather than a single narrow command, so treat it as a capability map: pick web, desktop, CLI, or MCP depending on whether a human designer or an agent owns the loop. Complexity is intermediate because local dev optionally requires Rust for Tauri.713installs106Pgmicro Postgres Sqlitepgmicro-postgres-sqlite is a Prism-oriented agent skill for solo builders who want PostgreSQL semantics without provisioning Postgres. It documents pgmicro, an in-process engine that compiles PostgreSQL SQL to SQLite VDBE bytecode through Turso, yielding fast embeddable storage in one file or memory. The skill covers npx and global CLI install paths, library embedding (including Rust-oriented triggers), wire-protocol mode for existing clients, and virtual system catalogs so tooling behaves like familiar Postgres. It suits local-first SaaS, offline agents, prototypes, and CI fixtures where spinning up Docker Postgres is heavy. Builders still need to validate compatibility for advanced PG features and production scaling limits compared to full server Postgres.712installs107Tribev2 Brain EncodingTRIBE v2 Brain Encoding is an agent skill that wraps Meta’s multimodal foundation model for predicting fMRI-like brain responses to naturalistic video, audio, and text. Solo builders and small research teams use it when they need in-silico neuroscience—encoding models, stimulus-driven cortical activity maps, or prototyping tribev2 inference and training without hand-rolling the full encoder stack. The skill documents pip install variants, HuggingFace preload, event dataframe construction from media files, and extension paths for visualization and Lightning-based training. It matters for agentic ML workflows where reproducible scripts and clear dependency groups beat one-off notebook glue, especially when you iterate on stimuli and model checkpoints frequently.711installs108Codex Plugin CcCodex Plugin for Claude Code bridges OpenAI Codex into your Claude Code session so solo builders can offload read-only reviews and longer investigative work to a second model without context-switching. Install it when you want a disciplined second opinion on diffs, a deliberate adversarial pass over tradeoffs, or a background subagent to chase bugs while you keep planning. The skill documents distinct entry points: standard review for uncommitted or branch changes, adversarial review when you need assumptions challenged, and rescue when you want Codex working asynchronously with polling via status and result. It targets indie teams already on Claude Code who also run the Codex CLI and Node toolchain. Treat it as agent orchestration glue—not a replacement for your own ship-phase security or test skills—so you still own merge decisions and secrets hygiene.708installs109Polymarket Arbitrage BotPolymarket Arbitrage Bot is an agent skill that packages a TypeScript automation for Polymarket’s short-horizon prediction markets. Solo builders use it when they want to implement the dump-and-hedge pattern: detect a sharp move on one side of a 15-minute Up/Down market, buy the dipped outcome, then hedge the opposite leg when the paired cost drops below a configured edge. The skill documents cloning the repo, installing dependencies, building to dist, and copying .env.example before tuning credentials and risk parameters. It emphasizes simulation-first operation so you can validate behavior without placing real CLOB orders, then promote to production only when PRODUCTION is explicitly enabled. The audience is indie builders comfortable with Node tooling who treat prediction-market automation as experimental software with real financial risk, not a turnkey investment product.704installs110Taiwan Equity Research CoverageTaiwan Equity Research Coverage is an agent skill that operationalizes the My-TW-Coverage repository: a structured equity database for 1,735 Taiwan-listed companies grouped into 99 sectors, each with business overview, supply chain links, customer and supplier relationships, and financial fields wired together by more than 4,900 wikilinks. Solo builders and small research teams install it when they need repeatable thematic search (“find names for a buzzword”), ecosystem tracing, valuation multiple refreshes, or disciplined additions of new tickers instead of ad-hoc spreadsheets. Invocation aligns with triggers such as building wikilink network graphs, generating Taiwan thematic investment screens, and auditing report quality. It is research infrastructure rather than a trading execution bot: clone the repo, install Python dependencies, and use agent-guided workflows over Pilot_Reports. Best paired with agents that can navigate markdown wikilinks, run data refresh scripts, and keep sector folders consistent as coverage expands.702installs111Torchcode Pytorch Interview Practicetorchcode-pytorch-interview-practice packages the TorchCode environment for solo ML builders who want interview-grade rehearsal: implement operators such as softmax, LayerNorm, multi-head attention, and GPT-2-related blocks from scratch, then submit to an automated judge for correctness, gradients, and runtime. The skill documents four onboarding paths—hosted Spaces, Colab with pip install torch-judge, Docker on ghcr.io/duoan/torchcode, or local clone—so you can practice without standing up a full training pipeline. With 40 curated problems, it mirrors LeetCode’s feedback loop for tensors rather than arrays. Use it when triggers mention torchcode, PyTorch coding interviews, or checking a scratch implementation; it is learning and validation tooling, not a deployment or observability skill. Indie builders sharpening skills before hiring loops or before writing custom layers in an agent-built model benefit most.702installs112Polymarket Copy Trading BotPolymarket Copy Trading Bot is an agent skill for solo builders who want to automate mirroring another Polymarket wallet’s BUY activity on Polygon mainnet. It documents cloning the TypeScript project, wiring environment variables for the CLOB SDK, choosing signature type for EOA or proxy wallets, and running continuous monitoring through REST polling and optional WebSocket feeds. Copy sizing flows through a configurable position multiplier with hard caps at session and market level, while order execution supports FOK, FAK, and LIMIT styles appropriate to liquidity. The skill emphasizes operational setup—triggers like copy trade polymarket wallet and polymarket bot environment setup—rather than investment advice. Builders should treat mainnet keys and API credentials as high risk and test thresholds before leaving the process unattended. It complements research skills by moving from observation to automated execution, but only for users who accept prediction-market regulatory and financial exposure.700installs113Codex Session PatcherCodex Session Patcher is a Python utility and web dashboard for solo builders who lose hours when a coding agent stops cooperating mid-session and the serialized transcript blocks resume. It scans Codex CLI and Claude Code JSONL logs plus OpenCode SQLite stores, detects refusal-shaped assistant turns, and replaces them with cooperative placeholders so native resume flows work again. A separate injection path adds security-testing and CTF-style system context into Codex profiles, Claude workspaces, or OpenCode workspaces when you are doing authorized lab work and want fewer policy dead-ends up front. The SKILL.md documents install via git clone, platform support tables, and triggers such as cleaning refusals or fixing Claude Code sessions after a block. Treat injection as high-trust configuration you own—appropriate only for environments you control and rules you are allowed to test.695installs114Learn Docker K8s GameLearn Docker & K8s Game is an interactive agent skill that turns your editor into a DevOps classroom: you say let's play, the agent reads AGENTS.md and engine rules, adopts Sarah the senior mentor, and walks you through broken-staging style incidents using your actual shell. Solo builders who dread another passive video course get spaced challenges, narrative stakes, and verify.sh gates so progress is proven on your machine—not in a simulated browser pane. It spans container basics through Kubernetes framing, which supports later Build containerization, Ship readiness, and Operate cluster work even though Prism shelves it under Operate infra as the primary home for runtime and platform literacy. Complexity is beginner-friendly assuming you can open a terminal in your repo clone; the skill does expect you to run real docker and kubectl style commands as chapters advance. Because it is markdown-driven open source from ara.so's trending collection, you can fork scenarios for your stack, but you should still treat production clusters carefully and never paste secrets into the story prompts. Ideal when you are onboarding to containers before shipping your first API or when you need a st694installs115Claude Better CliClaude Better CLI documents a compatibility-first reimplementation of the Claude CLI aimed at solo builders who live in the terminal. The skill explains measured wins such as faster `--help` and chat bootstrap times plus much lower resident memory during sustained sessions, while asserting near-perfect command and output parity so existing aliases keep working. Triggers cover explicit requests for a faster Claude CLI, memory tuning, or a drop-in replacement under the claude-better name. Availability is limited: upstream source may require maintainer access, so treat install steps as conditional on your onboarding package. Use it when agent startup latency or RAM spikes block flow across build, ship, and operate tasks—not when you need new Claude features unrelated to the harness. Prism lists it so builders can discover performance-oriented CLI options before committing migration.693installs116Ministack Aws EmulatorMiniStack AWS Emulator is an agent skill that teaches solo builders how to stand up a free, open-source local AWS-compatible gateway when cloud bills or account setup block iteration. It positions MiniStack as a LocalStack replacement that exposes many common services—storage, queues, functions, identity, parameters, email, and more—behind one port so existing boto3 and infrastructure-as-code flows need minimal rewiring. The README walks through pip and Docker install paths, optional port overrides via GATEWAY_PORT, and a quick health curl to confirm the stack is alive. That makes it valuable during Build while you prototype Lambdas or DynamoDB access, during Ship when you want integration tests in CI without touching real AWS, and during Validate when you prove an architecture locally before committing spend. It is not a substitute for AWS parity testing for every edge case or for production hardening; it is a fast mock plane for indie APIs and agent-generated IaC. Keep credentials fake, point SDKs at localhost, and treat service coverage as emulator-specific.693installs117Wildworld DatasetWildWorld Dataset is a research-oriented agent skill that packages what solo ML builders and small labs need to know about a massive action-conditioned game world dataset before writing training code. WildWorld targets dynamic world models—systems that predict future game visuals and state from history plus player actions—sourced from automated collection in a photorealistic ARPG. The skill surfaces headline scale (108M+ frames, hundreds of actions) and the annotation contract on every frame: skeletons, action and state vectors, camera extrinsics, and depth, which matters when choosing model architecture and loss design. It also maps entity diversity (monsters, weapons, stages) and points evaluators toward WildBench for standardized comparison. Triggers in the SKILL.md mirror common user phrases about loading WildWorld, ARPG annotations, and action-conditioned video research. It does not replace downloading or licensing the dataset; it gives your coding agent consistent procedural context so Validate and Build steps start from the same facts about labels and benchmark naming.690installs118Phantom Ai Coworkerphantom-ai-coworker is a setup and architecture skill for builders who want an AI teammate that behaves like staff on a real computer, not a browser tab that forgets yesterday. It documents how Phantom combines the Claude Agent SDK with TypeScript on Bun, vector memory in Qdrant, and a dynamic MCP server that can register new tools while running. Persistent memory and self-evolution hooks are framed so solo founders can delegate recurring ops, integrations, and light platform building on the same VM. Communication surfaces include Slack, email, Telegram, and webhooks, which matters when you are the only human on call but still need auditable async handoffs. The skill’s triggers cover Docker deploy, MCP server setup, Slack bot identity, and configuring the self-evolving loop. Expect advanced wiring: secrets, networking, and long-running processes. Use it when you are ready to invest in an agent-shaped product or internal co-worker rather than a single-session coding assistant.679installs119Career Ops Job SearchCareer-Ops Job Search is an agent skill package that wires Claude Code into an end-to-end hiring workflow: profile and portal YAML config, Markdown CV as the grounding document, and npm/Playwright for PDF output. Solo and indie builders use it when they need structured job hunting without juggling spreadsheets—evaluate fit before applying, batch-process listings, and keep scans and submissions visible in one dashboard. It is opinionated around the career-ops repo layout (config/profile.yml, portals.yml, cv.md) and Claude agents rather than a generic resume site. Install clones the GitHub repo, installs Node dependencies, and copies example configs before invoking the documented skill modes.674installs120Kordoc Korean Document Parserkordoc-korean-document-parser documents how solo builders use the kordoc TypeScript library and CLI to read Korean government and office formats—HWP, HWPX, and PDF—and emit Markdown and IRBlock[] data agents can reason over. It matters when your content lives outside plain text: procurement specs, forms, and legacy HWP archives that Cursor or Claude cannot open natively. The skill walks installation (npm install kordoc, optional pdfjs-dist), the parse(buffer) API with ArrayBuffer input, and operational triggers such as diffing two HWP files or extracting form fields. You can run one-off conversions from the terminal or expose parsing through an MCP server so coding agents fetch structured text during Build integrations. Intermediate complexity assumes comfort with Node, buffers, and optional peer dependencies. It is a task integration skill, not a journey-wide methodology—pick it when Korean document conversion is on your critical path.673installs121Caveman Token OptimizerCaveman Token Optimizer is an ara.so agent skill that forces compressed caveman-speak responses in Claude Code and Codex so solo builders pay for fewer output tokens on long coding sessions. It removes social filler and verbose transitions but leaves code, APIs, and error strings intact. You install it when context windows fill fast from chatty defaults, not when you need polished prose for customers. Triggers cover npx install, marketplace plugin setup, and plain-language requests to reduce token usage. It is a behavioral overlay, not a linter or planner—pair it with normal review skills before you ship user-facing copy. Works across whatever phase you are in because the savings apply to every agent turn once enabled.672installs122Claw Code HarnessClaw Code Harness is an agent skill for solo builders who fork or replace the Claude Code harness with a Python (and emerging Rust) implementation and need visibility into what was ported. It centers on a CLI entrypoint and importable modules for manifest summaries, subsystem lists, parity audits versus archived upstream behavior, and tool/command metadata inspection. Use it when you are maintaining a custom harness, comparing ports, or documenting which tools and commands your agent runtime exposes. There is no PyPI package yet—install from the instructkr/claw-code repository and run unit tests to confirm the tree. It does not replace day-to-day coding; it makes harness maintenance auditable so your agent stack stays trustworthy before you ship workflows that depend on specific tools.670installs123Clawd Code Python Portclawd-code-python-port is a reference skill for solo builders who want to understand and run clawd-code, a from-scratch Python port of the Claude Code agent harness. It explains what the project does—tool registration, command dispatch, task orchestration, and CLI startup—and how to install the repo, optionally use a virtualenv, and drive work through oh-my-codex on top of OpenAI Codex. Use it when you are building or extending your own agent CLI, comparing behavior to Claude Code concepts, or running parity and workspace checks after local changes. The audience is developers comfortable reading Python project layout and agent patterns, not teams looking for an official Anthropic product. It is educational and third-party; treat licensing and API keys as your responsibility. Pair with implementation skills once you know which module or command you are changing.665installs124Bazi Skill Chinese AstrologyBaZi Skill — Chinese Astrology is a niche Claude Code skill that turns birth data into a traditional Four Pillars chart and narrative destiny reading grounded in nine classical references. Solo builders shipping astrology, wellness, or culturally localized agent products can study it as a worked example of multi-step data collection, domain rules, and bilingual triggers rather than as generic SaaS scaffolding. End users invoke it when they want structured 八字排盘 and interpretive copy in the same session. The skill emphasizes interactive prompts, pillar derivation, cycle tables, and themed life guidance instead of generic LLM fortune-telling. It fits builders learning how to package deep vertical knowledge into SKILL.md with explicit invoke triggers. It is not a substitute for licensed divination services, medical advice, or rigorous demographic analytics for mainstream B2B apps unless you add your own compliance layer.663installs125See Through Anime Layer DecompositionSee-through is an agent skill that packages expertise for decomposing a single anime illustration into manipulatable 2.5D assets. Solo and indie builders who animate characters, build illustrated games, or pipeline VTuber art install it when they need semantic layer separation—not manual cutouts—from one source image. The framework uses diffusion models to inpaint occluded regions so each of up to twenty-three layers is complete, infers drawing order with a fine-tuned Marigold depth model, and exports a layered PSD with masks and depth maps. Triggers align with tasks like splitting an illustration into PSD layers, extracting body parts, or building a 2.5D model from a single frame. Expect a Conda Python 3.12 environment, PyTorch stack, and GPU-friendly setup rather than a one-click SaaS. Use it during build when layered PSDs unblock rigging, parallax, or compositing; skip it for non-anime styles or when flat PNGs are enough.648installs126Open Agent SdkOpen Agent SDK teaches solo builders to run autonomous agent loops inside TypeScript or Node.js using `@shipany/open-agent-sdk`, positioned as a fully open-source, in-process counterpart to Anthropic’s Claude Agent SDK. The engine runs in your process—no spawned Claude Code CLI—so the same patterns work on cloud VMs, serverless handlers, containers, and pipelines. Installation is a standard npm add with Node 18+; authentication flows through `ANTHROPIC_API_KEY` or third-party gateways such as OpenRouter via `ANTHROPIC_BASE_URL` and model overrides. Use when you are wiring agent orchestration into a product backend, replacing CLI-coupled SDK assumptions, or deploying agents where subprocess tooling is blocked. It assumes you bring TypeScript project structure and API keys; it delivers configuration and usage patterns for durable in-process loops rather than desktop-only workflows.647installs127Emdash CmsEmDash CMS is an agent skill for solo builders shipping content-driven sites on Astro who want a WordPress-like extensibility model without PHP monoliths. It walks through scaffolding with npm create emdash@latest, choosing templates and Cloudflare or Node.js backends, adding the package to an existing Astro project, defining content types and collections, querying Portable Text data, building sandboxed plugins, configuring Cloudflare D1 and R2, and wiring passkey-first auth. Use it when your agent should own the CMS integration instead of hand-rolling markdown folders or coupling to a separate headless vendor. It matters for indie sites, marketing pages, and blogs where you want structured content, developer ergonomics, and edge deployment on Workers. Complexity sits at intermediate: you need comfort with Astro, TypeScript, and either Cloudflare bindings or local SQLite.621installs128Openclaude Multi LlmOpenClaude Multi-LLM is an agent skill for solo builders who want Claude Code ergonomics without locking inference to a single vendor model. It explains OpenClaude, a fork that sends chat completions through openaiShim.ts so any provider speaking the OpenAI API can sit behind the same agent tools you already rely on for shipping code. The skill covers npm global install, Bun-based source builds, and quick dev runs, then focuses on the activation step: setting the environment variables that select base URL, credentials, and model name. That makes it practical when you are comparing cost and quality across GPT-4o, local Ollama, DeepSeek, or cloud Gemini routes while keeping MCP and filesystem workflows intact. Use it during Build whenever you are standardizing one CLI for multiple models or dogfooding a self-hosted endpoint before production agent features.621installs129Cc Gateway Ai ProxyCC Gateway AI Proxy helps solo builders and small teams run Claude Code behind a privacy-oriented reverse proxy instead of sending raw device fingerprints and scattered OAuth state from every laptop. The skill explains architecture: clients hit CC Gateway, which rewrites headers, environment-derived signals, and system-prompt-related identity dimensions before forwarding to api.anthropic.com, while token refresh talks only to platform.claude.com. It pairs operational guidance with triggers like Docker setup, multi-machine gateways, and disabling side channels. You reach for it during build when integrating agent tooling with your network policy, again at ship when you want consistent identity before wider rollout, and at operate when you maintain proxy config and Clash rules. It is intermediate complexity because you must coordinate env vars, network rules, and gateway credentials without breaking the agent loop.618installs130Open Multi Agent OrchestrationOpen Multi-Agent Orchestration teaches solo builders how to use the open-multi-agent TypeScript framework to run production-oriented agent teams: distinct roles, models, and tools collaborating toward a shared goal. The orchestrator resolves task dependencies as a DAG, runs work in parallel where safe, and coordinates shared memory and inter-agent messaging inside one process—avoiding brittle subprocess fan-out. Triggers in the skill mirror real prompts: set up a multi-agent team, pipeline tasks with dependencies, run agents in parallel, or blend Claude and GPT in a single workflow. Install via npm or pnpm, set ANTHROPIC_API_KEY and optional OPENAI_API_KEY, then define teams, custom tools, and schedules. It fits indie builders who have outgrown single-threaded agent chats but do not yet want a separate orchestration SaaS. Expect intermediate TypeScript comfort and clear task graphs before you adopt the pattern.612installs131Claude Code Source RecoveryClaude Code Source Recovery is an agent skill for solo and indie builders who want to learn from Anthropic’s accidentally published Claude Code 2.1.88 source map rather than guessing from minified JavaScript. It orients you across entrypoints, the command system, MCP integration, and Ink/React terminal UI modules so you can compare your own agent CLI design to a production implementation. Use it when you need to answer how commands are dispatched, how MCP is wired, or how session and tool execution are structured before you commit to an architecture. It is research and reference work—not a shortcut to fork proprietary code—best paired with your own compliance review of licenses and redistribution rules. Triggers include exploring recovered source, understanding CLI internals, and studying codebase structure after map extraction.598installs132Pretext Text MeasurementPretext Text Measurement is an agent skill documenting @chenglou/pretext, a JavaScript/TypeScript library for fast, accurate text sizing and wrapping without touching the DOM. Solo builders shipping custom editors, canvas dashboards, Figma-like tools, or any interface that must know line breaks and heights before paint reach for it when browser reflow would tank performance. The workflow centers on caching prepare() results that hit the real font engine once, then calling layout variants on every width change with pure arithmetic. It supports plain paragraphs, segmented rich text, shrink-wrapped widths, and wrapping around images per the skill triggers. Install with npm install @chenglou/pretext and wire it where React or CSS cannot give deterministic pre-layout metrics. Intermediate familiarity with canvas fonts and TS modules is enough; you still own accessibility and fallback DOM measurement for screen readers where needed.592installs133Sheets Terminal SpreadsheetSheets Terminal Spreadsheet is an agent skill for invoking the Go TUI `sheets` so you view and edit CSV files entirely in the terminal with vim-like keybindings. It targets indie builders and operators who want spreadsheet workflows without context-switching to GUI apps—budget lines, exports from scripts, or quick pipeline dumps piped into `sheets`. Use it when triggers match editing CSV in terminal, opening a spreadsheet from the command line, or needing maaslalani sheets behavior during local data cleanup. The skill documents installation, launching the TUI, stdin and pipe patterns, and non-interactive cell or range reads for scripting. It matters for agent-assisted dev sessions where the model should recommend the right CLI instead of hallucinating Excel steps.571installs134Omnivoice TtsOmniVoice TTS is an expert agent skill for wiring OmniVoice—a massively multilingual zero-shot text-to-speech stack—into solo-builder projects. It walks through environment setup (Python 3.9+, PyTorch 2.8+, CUDA, Apple MPS, or CPU), pip and source installs, and the core capabilities builders care about: cloning a speaker from short reference audio, designing voices with descriptive attributes, and running inference at scale including batch jobs. The skill positions OmniVoice as a practical alternative to hand-picking per-locale voices when you need broad language coverage or rapid voice prototyping for agents, narrated content, accessibility, or demo apps. Expect GPU-backed workflows for serious throughput; the docs emphasize platform-specific torch wheels. Use it when your agent should generate correct install commands, choose inference modes, and avoid common PyTorch/CUDA mismatches—not when you only need a hosted SaaS TTS API with no local model.566installs135Metatron Pentest AssistantMETATRON Penetration Testing Assistant is an agent skill for advanced solo builders and security-minded developers who want AI-guided recon and reporting entirely on their own machine. It targets Parrot OS Linux with a modular layout: metatron.py for menus and scan orchestration, db.py for MariaDB CRUD across history and vulnerability tables, tools.py wrapping standard recon binaries, llm.py for Ollama-driven analysis and tool dispatch, and search.py for DuckDuckGo and CVE context without paid search APIs. Every scan gets a serial identifier in the database so you can query history, track fixes and exploits, and export consolidated findings to PDF or HTML. Triggers cover initial setup with Ollama and MariaDB, automated runs, custom recon extensions, troubleshooting model or DB issues, and historical queries. This is operator-grade tooling—not a substitute for professional certification or legal authorization—and fits the ship phase when you are hardening something you are about to expose.553installs136Tong Jincheng Relationship SkillTong Jincheng Relationship Skill is a Claude Code agent skill that applies a distilled “深情祖师爷” interpersonal framework to relationship doubts, trust tests, and balancing early-stage work with romance. Solo and indie builders who ship alone still lose weeks to ambiguous dating dynamics and unspoken needs; this skill routes those questions through nine explicit heuristics (uncertainty as disinterest, don’t test humanity—state needs, attraction over讨好) so the agent answers with consistent judgment instead of generic empathy. It is not a coding or growth-hacking tool—it is persona-guided coaching for decisions that affect focus during validate, build, and grow. Use it when triggers match and you want structured pushback before you over-invest time or emotional bandwidth. The catalog lists it under Grow for lifecycle framing, but invoke whenever relationship noise blocks execution.540installs137Parlor On Device AiParlor On-Device AI teaches solo builders how to run a real-time multimodal assistant entirely on their own hardware. A browser captures microphone audio and camera frames, streams them over WebSocket to a FastAPI server, and plays back synthesized speech while showing transcripts. Understanding uses Gemma 4 E2B through LiteRT-LM on GPU; speech output uses Kokoro TTS with a platform-aware backend (MLX on Mac, ONNX on Linux). Silero VAD in the browser enables hands-free turns and barge-in so users can interrupt mid-utterance. The architecture is aimed at privacy-sensitive demos, offline-capable copilots, and Apple Silicon laptops where cloud APIs are undesirable. Triggers cover setup phrases like configuring the Parlor websocket voice-vision server or running local voice AI with camera. Expect intermediate complexity: model weights, GPU paths, and WebSocket protocol tuning. After integration, you typically harden ship-phase testing and perf before any public launch.536installs138Free Code Claude Clifree-code Claude CLI is an agent skill for solo builders who want a terminal-native coding agent based on Claude Code without default telemetry and with experimental features enabled at build time. It walks through the recommended curl installer, a manual Bun build from the paoloanzn/free-code repository, and the mental model of what differs from the stock npm release: fewer injected guardrails, more feature flags, and explicit multi-cloud model routing. Use it when you are standardizing your local agent stack, comparing providers (Anthropic, OpenAI Codex, Bedrock, Vertex, Foundry), or reproducing a known-good CLI from source rather than trusting a black-box binary. Because it modifies how your primary coding agent behaves and connects to cloud APIs, treat API keys, install scripts, and fork updates as part of your normal security hygiene.534installs139Fieldtheory Cli BookmarksField Theory CLI (`ft`) is a global npm CLI that downloads your X/Twitter bookmarks locally, indexes them for search, optionally classifies them with an LLM, and exposes everything through terminal commands your coding agent can run. Solo and indie builders use it to turn a noisy save-for-later pile into a queryable research library—especially when triggers like syncing bookmarks, searching saved posts, or feeding bookmarks into Claude Code match the moment. Installation is `npm install -g fieldtheory` with Node 20+, Chrome logged into X on macOS for default sync, or OAuth elsewhere. Quick start is `ft sync`, then `ft search` or `ft viz`. It matters for Prism because it is a concrete CLI integration in the agent-tooling and research lane, not a generic chat prompt.532installs140Paper2code Arxiv Implementationpaper2code — Arxiv Paper to Working Implementation is an agent skill for solo builders and indie ML hackers who want reproducible code from primary literature, not a hallucinated notebook. You invoke it with an arXiv URL—optionally forcing JAX, PyTorch, or another framework—when you need a working Python baseline for an experiment, benchmark, or product feature grounded in the paper's notation. The skill's contract is strict provenance: generated logic should reference the implementing section or equation, and anything underspecified in the PDF surfaces as an audited ambiguity rather than a silent default. That makes it useful for Validate-style spikes that graduate into Build implementations, teaching workflows, and competitive reproduction efforts where reviewers ask "which equation is this?" Prism catalogs it for agents that otherwise summarize papers without shipping testable code, complementing generic coding skills with a research-to-repo ritual tuned to arXiv links and the /paper2code command surface documented in the skill.531installs141Tong Jincheng Interpersonal SkillTong Jincheng Interpersonal Skill injects a documented interpersonal cognition framework into Claude Code–style agents so builders can stress-test dating, negotiation, or team dynamics questions with consistent principles. The package ships SKILL.md plus six research references; triggers include 童锦程, 深情祖师爷, and 景辰怎么看. It targets solo operators who want blunt human-nature heuristics when drafting user interviews, support replies, or founder conversations—not entertainment cosplay. Use journey-wide whenever you need a second lens before sending a message or changing positioning; skip when you need legal, clinical, or harassment-sensitive mediation where professional standards apply.520installs142Claude Code Source StudyClaude Code Source Study is a research-oriented agent skill that routes you through a structured source-analysis curriculum for Anthropic’s Claude Code CLI—roughly nineteen hundred files of Bun, TypeScript, Ink terminal UI, and Anthropic API integration. Indie builders who want production-grade agent architecture rather than toy demos use it when studying how system prompts, tool registration, permission gates, and orchestration actually ship in a paid product. The skill is meta: it does not run Claude Code for you; it teaches how to read the repo module by module using the bundled article index and triggers such as learning AI agent architecture or understanding the internal tool system. Expect intermediate-to-advanced depth, bilingual context (series in Chinese with technical anchors), and patterns you can transplant into Cursor, Codex, or your own CLI agent. Best after you have a repo to experiment in and before you commit to a custom agent stack.515installs143Cabinet Ai Knowledge BaseCabinet AI Knowledge Base documents how to install and run Cabinet, an AI-first startup OS from the ara.so Daily 2026 collection where notes, agents, and apps live as markdown on disk. Solo and indie builders use it when they want a self-hosted alternative to Notion-plus-chat silos: optional KB_PASSWORD auth, DOMAIN configuration, Claude Code CLI as a prerequisite, and dev:all serving the web UI with wizard-driven team setup. The skill covers quick start via npx create-cabinet, manual clone paths, and environment variables so you can add agents, schedule jobs, and embed lightweight tools without surrendering file ownership. It spans validation notes, build-time specs, launch checklists, and day-two operations because the same workspace holds research, plans, and runbooks—ideal when you are building an AI team rather than bolting a single API into an existing repo.511installs144Awesome Design Md JpAwesome Design MD JP helps solo builders ship Japanese web products without agents guessing typography. It extends the Stitch DESIGN.md pattern with explicit rules agents can follow: font stacks that fail gracefully across macOS and Windows Japanese faces, looser line-height than Western defaults, kinsoku-driven line breaks, and mixed Japanese–English composition. Invoke it when you are creating or extending DESIGN.md for a JP landing page, dashboard, or marketing site and want one plain-text source of truth Codex, Cursor, or Claude Code reads before writing CSS or components. The skill is beginner-friendly on concept but intermediate in execution because you must align brand tokens with real font availability. Output is a DESIGN.md artifact, not automated screenshots.506installs145Karpathytalk CommunityKarpathyTalk Community teaches agents and solo builders how to run KarpathyTalk, a Go-based developer social network where posts are markdown and the API is openly readable by LLMs. It covers GitHub OAuth app creation, local configuration, deployment to a server, and programmatic posting or ingestion for automation. The stack is intentionally small—htmx, goldmark, SQLite—so an indie operator can own the full social layer without managed SaaS lock-in. Use it when you want a Twitter-meets-Gists forum for your community, a public feed your coding agent monitors, or a self-hosted discussion hub alongside your OSS projects. Intermediate complexity reflects OAuth, env vars, and production deploy steps.505installs146Binance Futures Signal BotBinance Futures Signal Bot is a Daily 2026 Skills package from ara.so that walks solo builders through cloning, installing requirements, and configuring an automated futures trader across Binance, Bybit, and OKX. It targets the painful middle ground between discretionary signals and disciplined execution: parse incoming alerts, size positions, set leverage, manage trailing stops, and watch P&L in a Rich-powered terminal UI. Triggers in the skill mirror real support queries—TradingView webhook wiring, leverage tuning, and bot-not-trading troubleshooting—so agents invoke it when the user names those exchanges or signal sources. You need Python 3.10+, API keys with futures permissions, and comfort with capital risk; the skill educates on setup, not financial advice. Tag it Build integrations for the wiring work, with secondary Operate monitoring once the bot runs 24/7. It complements generic backend skills but is explicitly trading-domain: pandas OHLCV and ta indicators support strategy extensions inside the same repo. Document for AEO that safety is paramount—users must audit keys, testnet behavior, and exchange rules outside the agent chat.471installs