Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Discover your DISC personality type and install personalized AI communication skills. Use when someone mentions personality, communication style, DISC, how AI should talk to them, how to communicate with a coworker/boss/teammate, preparing for a meeting or difficult conversation, or wants to customize their AI experience. Also triggers on: 'what's my personality type', 'set up my personality', 'personality quiz', 'how should I talk to', 'communication coaching'. Includes interactive quizzes for both flows. Powered by Crystal's DISC framework.
Validate skills against agentskills.io specification. Use when adding new skills to the marketplace, reviewing skill PRs, checking skill compliance, or running quality gates on skills. Validates frontmatter fields (name, description, compatibility, metadata, allowed-tools), directory naming, line limits, and structure.
Anti-bot detection toolkit for AI agents. Diagnoses why your agent is getting blocked and fixes it. Use when: agent hits 403/429 errors, encounters CAPTCHAs (hCaptcha, reCAPTCHA, Cloudflare Turnstile), gets detected as a bot, receives fake/empty data from websites, encounters Cloudflare challenges, needs to bypass IP blocks, or is running on a datacenter/cloud IP. Covers: residential proxy setup, CAPTCHA solver configuration, stealth browser fingerprinting. For web scraping, browser automation, form filling, and any task where bot detection is a problem. Compatible with Puppeteer, Playwright, Selenium, and headless Chrome.
Coordinate multi-agent swarm workflows. Use when working in parallel with other agents, managing shared resources, or orchestrating distributed tasks. Covers conflict prevention, handoffs, and two-tier task tracking (native task list + artifacts).
List, select, and install Volcengine skills from the volcengine-skills marketplace. Use when a user asks which Volcengine skill or plugin handles a task, wants to browse every available skill, needs an optional skill that is not currently installed, or asks to install or check the installation status of a Volcengine skill. Also use to discover any core skill, including this finder. Use as a fallback during a Volcengine operation when the currently loaded or installed skills cannot cover the required product, tool, or workflow; inspect the full catalog, select the missing capability, and install it.
Dual-team project workflow — Builder + Observer teams, cost tracking, parallel execution, security gates, agent orchestration. Use when: start day, begin session, status check, new feature, build, implement, end day, wrap up, debug, investigate, dual team, observer team, builder team, spawn observers, devil's advocate.
ADHD Progress Hub continuity protocol for substantial coding work. Use the operator's adhd-hub MCP when starting/resuming meaningful project work, pausing incomplete work, finishing a tracked task, or setting a reminder. Use the forge mailbox when Hub MCP is unavailable to a remote agent. Do not use for trivial questions, read-only lookups, or tiny edits.
Carry requested work through authorized completion with engineering playbooks, relevant installed skills and useful workers. Use as the operating entrypoint, for a named playbook, to compose or resume work across owners, resolve the next route, or request the installed skill inventory. A settled single-owner task can use its specialist directly.
Render independent questions from an Agent workflow as a local interactive form, preselect recommended answers, collect an optional supplement for every question, save responses as portable JSON, and return submitted answers directly to the waiting Agent command. Use for grill-me, grill-with-docs, brainstorming, requirement clarification, configuration, planning, or any other workflow that needs user confirmation or question collection; explicitly invoke Ask UI when one round contains more than two questions. Also use the manual recovery path when the user says “已提交”, “提交好了”, or “答完了” after an active Ask UI round.
Install skills from the AVIZ Skills Library. Use when user wants to install a skill, browse available skills, set up Claude Code skills, or asks about skill installation. Triggers on "install skill", "add skill", "setup skill", "available skills", "skill library", "browse skills".
Hub MỘT TÊN cho dây chuyền sản xuất kiểu Ralph (GH#15): biến tài liệu thô của user thành sản phẩm qua 5 mode — `/br interview` (đọc folder raw/ → đối chiếu bộ specs chuẩn S1–S10 → sinh bộ câu hỏi hỏi-bù dạng HTML xem trước + MD điền, cho phép lens chuyên gia điền thay mục chưa chắc), `/br compile` (answers → BR.md có clause_id + bảng "Giả định đang gánh"), `/br slice` (BR → frames nhỏ gắn chặt code, gác bằng frame-lint), `/br run <frame>` (mỗi frame chạy loop-runner có 6 phanh + dry-run + người gác · sau frame xanh TỰ chạy QC tất định 0-token), `/br qc` (audit senior 4-mục: /qc-code soi code + /qc-uiux soi UI/UX qua engine visual-qa), `/br status` (trang line-status.html tất định: frame nào chạy/kẹt/xong + truy ngược lỗi→frame→clause). Gọi khi user nói "br", "interview", "phỏng vấn yêu cầu", "soạn BR", "slice frame", "chạy frame", "qc frame", "audit ui mockup", "trạng thái dây chuyền", "ralph pipeline", hoặc invoke /br. User chỉ mô tả PHẠM VI, không cần nhớ lệnh con.
Enters idea-maturation mode — researches, analyzes critically, points out trade-offs and suggests improvements without creating or executing anything until the user explicitly asks. Use when the user wants to explore, discuss or refine an idea before implementing it. Triggered by phrases like "quero criar", "tenho uma ideia", "pensa comigo", "me ajuda a planejar", "vamos fazer um brainstorm", "entra em modo brainstorm", or any product, feature, system, skill or workflow idea not yet specified enough for implementation. Has a specialized lens for development project specification — also activate when the user wants to "especificar um projeto", "levantar os requisitos", "montar a spec antes de codar", "definir a arquitetura antes de começar", or describe a system they intend to build. Also activates when the user asks to document, summarize or preserve in markdown a brainstorm already in progress or just concluded, regardless of topic (system, process, email, idea, argument, etc.), and when they want to resume a saved brainstorm ("continua o brainstorm de ontem", "retoma esse brainstorm", or pointing to a brainstorm-*.md file).
Analyze a repository to generate recommended Claude Code settings.json permissions. Use when setting up a new project, auditing existing settings, or determining which read-only bash commands to allow. Detects tech stack, build tools, and monorepo structure.
This skill should be used when the user asks to "create a team", "spawn agents", "assign tasks", "coordinate multiple agents", "check team status", "view kanban board", "send messages between agents", "manage team tasks", "monitor team progress", or mentions "clawteam", legacy "oh", "multi-agent coordination", "team collaboration", "agent inbox", "task board", "spawn worker". This skill should also be triggered when the current task is complex enough to benefit from splitting into subtasks and delegating to multiple agents — for example when the user asks to "build a full-stack app", "refactor the entire codebase", "implement multiple features in parallel", or when the agent determines that the work scope exceeds what a single agent can efficiently handle alone. Provides comprehensive guidance for using the ClawTeam CLI to orchestrate multi-agent teams with task management, messaging, monitoring, runtime profiles, git context, and recovery tooling.
This skill should be used when creating a skill for a CLI tool. Use when users ask to document a command-line tool, create CLI guidance, or build a skill for terminal commands. Essential for systematically introspecting CLI tools through help text, man pages, GitHub repos, and online research, then organizing findings into effective skill documentation.
Deploys Llama Guard 3 safety classification, NeMo Guardrails programmable dialogue rails, and LLM Guard input/output scanner pipelines as complementary runtime defenses that inspect and constrain LLM prompts and responses. Use when adding a production runtime safety layer to an LLM, RAG, or agent application to block jailbreaks, prompt injection (OWASP LLM01), toxic content, or sensitive-data leakage before it reaches or leaves the model.
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Ask another coding agent in a Herdr pane—on this machine or another—and get exactly one authorized reply, with a loop guard, depth limit of 1, a fan-out cap, and a record of every delegation; send one-way notices; keep a plan’s lead with epochs and succession; lead a multi-agent debate to decide, research the state of the art, stress-test, or verify (rounds, angles, red teams, counted ballots and claims, a panel seated from idle panes or agentkit accounts). Use only when the user mentions Herdr, peers, a debate, panel or research by several agents, another agent, pane or machine, or when a plan you are executing delegates a task through Herdr. Do not use merely because parallel work could help. Requires HERDR_ENV=1 and Herdr’s official herdr skill.
Initialize the memory system in the current directory, generating CLAUDE.md (optionally AGENT.md for Cursor), MEMORY.md and the memory/ directory. Triggered when the user says "initialize memory", "set up memory", "memory init" or "/memory-init".
Create, refine, or fix NVIDIA voice agents (Cascaded or Omni) with Pipecat or LiveKit. Use when building, scaffolding, or iterating on a real-time voice-agent pipeline, including speech (ASR/TTS) customization and cloud or local deployment. Not for offline/batch speech-to-text, text-only chat or RAG, or generic Docker or CUDA work unrelated to a voice agent
Use when the user asks to review code/a PR with opencode or with a non-Claude model (e.g., "review with opencode", "review using gpt-5.6-sol", "/opencode-review"). Runs a fresh opencode session in a subagent, tells it to use the review-code skill, and relays its review.
Configure and interpret the Phoenix plugin for Harbor agent evaluations. Use when adding `arize-phoenix` to Harbor jobs, choosing ATIF tracing, mapping Harbor tasks and rewards to Phoenix experiments, comparing agents or models, resuming jobs, or troubleshooting Harbor records in Phoenix.
PHP web full-chain white-box code security audit pipeline (multi-file orchestration). Serves as the master orchestration reference, calling each sub-audit skill by sink type (`php-route-mapper/php-auth-audit/php-route-tracer/php-*-audit`) and reusing unified output and severity standards.
Pi headless RPC JSONL over stdin/stdout, strict LF framing, --mode json event streaming, and Node SDK alternatives (AgentSession). Use when integrating pi --mode rpc, pi --mode json, writing a client that speaks the protocol, reading JSON event streams, or choosing RPC vs in-process SDK. Use for "pi rpc client", "JSONL framing", "readline rpc", "--mode json", "json events", "embed pi without subprocess", "createAgentSession", "AgentSessionRuntime", "ModelRegistry", "embed pi", "SDK factory", even if the user only pasted JSON lines.
Two writing rule sets for readers who may not read English fluently: COMMS for live agent-to-user text (interview questions, confirmations, blockers), ARTIFACT for persistent artifact prose (proposal.md, specs, design.md). A caller states which set applies. Evidence backing is in docs/research/2026-08-07-plain-language-non-native-readers.md.
This skill should be used when the user asks to "optimize a prompt", "improve prompt performance", "design a prompt template", "write better prompts", "debug prompt issues", "use chain-of-thought", "structured prompting", "few-shot prompting", or wants to apply advanced prompt engineering patterns for production LLM applications.
Deep recursive research with a self-regulating loop up to PhD level. Applicable to any domain (science, technology, business, art, humanities). Uses WDM + Munger inversion for autonomous decisions, trusted-source tiering, and disk checkpointing to survive context limits.
Run the reviewer's half of the document review loop without the user relaying each round — review the current <DOC>.revN.md into <DOC>.revN.notes.md, commit the closed round, wait for the author's rev(N+1), and repeat until the document is ready for final, a round cap is hit, or only the user can move it forward. Use after rev1 exists, when the user invokes /review-doc-loop <feature> [max-rounds] in a fresh session that did not write the document. Its counterpart is answer-notes-loop in the author's session
Create and manage Claude Code skills following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns, file paths, content patterns), enforcement levels (block, suggest, warn), hook mechanisms (UserPromptSubmit, PreToolUse), session tracking, and the 500-line rule.
Single-dimension execution and multi-source search integration within deep research; use when plan.json already defines dimensions and you need to call the appropriate search capabilities according to dimension.search_strategy for multiple rounds of retrieval, evidence screening, and cross-validation, then write to sub_reports/{dimension_id}.md.
Generates structured test assertions and failure diagnostics for skill packages from a definition and task prompt. Triggers on: "verify this skill", "generate assertions", "surrogate verification", "diagnose skill failure". NOT for code review, use pr-review.
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.