Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Audit one research Workspace for declared Unit outputs and Pipeline target Artifacts, writing `output/CONTRACT_REPORT.md`; use for mid-Run coverage snapshots or final delivery completeness, not deep provenance integrity.
Use when the user wants to stress-test a plan, design, migration, or architectural decision before building or committing; de-risk a major decision; examine design tradeoffs; assess whether a product or feature is worth building (problem evidence, demand test, go/no-go); or says “grill me.” It then interviews the user relentlessly, one question at a time, with a recommended answer for each, exploring the codebase before asking until every branch of the decision tree is resolved and there is shared understanding.
Manage an engineering backlog for humans and AI agents: classify issues, improve issue quality, sync PR state, and label safe tickets for agent execution. Use when GitHub Issues, GitHub Projects, Linear, or an explicit local backlog path is the source of truth.
Supplies the project-specific conventions, architecture, and contribution guidelines for the better-chatbot open-source AI chat platform (github.com/cgoinglove/better-chatbot; Next.js 15, Vercel AI SDK 5, Better Auth, Drizzle ORM, PostgreSQL). Covers the API architecture, the three-tier tool system (MCP, Workflow, Default), server action validators and the safe() wrapper, database repository patterns, compound component design, DAG workflow builder internals, Playwright E2E test setup, and principles such as progressive enhancement, defensive programming, and streaming-first, plus templates for adding tools, API routes, and repositories. Use when working in the better-chatbot repository, contributing features or fixes, adding API routes or database queries, implementing tools and workflows, building UI components, or integrating multiple AI providers.
bstack — the Broomva Stack. Twenty irreducible primitives (P1-P20) plus 30 curated skills that turn an agent-driven workspace into a self-operating system. Each primitive closes one specific failure mode (session amnesia, destructive ops, invisible work, dirty-tree drift, context rot, ritual phrases without substance, single-model echo chambers). The canonical operating mode is broomva/autonomous (fires every reflex without prompting). Use when: (1) bootstrapping a new agent-driven workspace, (2) verifying primitive compliance via 'bstack doctor', (3) repairing missing governance/hooks/policy via 'bstack repair', (4) listing installed-vs-missing skills via 'bstack status', (5) validating skill frontmatter health via 'bstack validate', (6) full reconfiguration via 'bstack revamp'. NOT FOR: incident or ticket severity labels — a 'P1 outage' or a P2 bug is not primitive P1 or P2; scaffolding an app or framework (bootstrap here is the governance metalayer, not a Next.js or Cargo template); editing bstack's own scripts, tests or docs (ordinary repo work on the bstack product, not an invocation of it); or a passing mention of bstack in a request whose imperative names a different skill. Triggers on 'bstack', 'broomva stack', 'bootstrap project', 'install all skills', 'P1' through 'P20', 'agent harness', 'self-operating workspace'.
Builds robust Python MCP (Model Context Protocol) servers with FastMCP—covering tool design, error contracts, event-loop-safe blocking work, subprocess/CLI wrapping, single-file vs. packaged distribution, global-state-free testing, and prompt-injection awareness. Use when writing an MCP server, exposing a tool or CLI to an LLM client, debugging tool registration or packaging, or testing MCP tools.
Ultra-compressed communication mode. Cuts ~75% of token usage by dropping filler, articles and formalities while keeping full technical accuracy. Use when the user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.
Must read before using the hosted ChatCut plugin in Codex, including creating projects, opening the editor, or editing/creating videos. Also use when ChatCut tools are missing or need authentication. For ChatCut Desktop, follow its own instructions unless the user explicitly chooses the hosted plugin/web surface.
Canonical joelclaw mail coordination protocol for all agents. Use when announcing work, checking inboxes, reserving files, releasing locks, handling task handoffs, triaging stale reservations, or auditing coordination quality. Triggers on: 'clawmail', 'agent mail', 'joelclaw mail', 'reserve files', 'release locks', 'handoff', 'agent coordination', 'mail status', or any multi-agent collaboration workflow.
Explore codebases efficiently with code2prompt, using sem-backed entity maps and grouped source reads to reduce agent round trips. Use when understanding an unfamiliar repository, locating implementations, gathering context for changes or reviews, or installing code2prompt
Сотрудничество делателей поверх контуров — агент передаёт работу другому и получает ответ. Триггеры: «спроси другого агента», «передай задачу», «это не моя зона», «эскалируй владельцу», collaborate, ask another agent, escalate to the owner, кадр на канале. Граф — запись, канал — побудка. Слышать ответ требует исполненного standing; говорить — нет, и первое слово идёт раньше. Холодный старт: «нет моей роли», «некому адресоваться», «написать своему человеку». Композирует standing, writing, inquiry, entry. Нужны iskron_*.
Run any question, plan, PR, or code through a Diverse Multi-Agent Debate (DMAD) council of 5 AI advisors with distinct reasoning methods. Advisors collaborate, peer-review each other anonymously, and a chairman synthesizes a verdict. Empirically outperforms adversarial debate (M3MADBench 2026, DMAD ICLR 2025). Use when: 'council this', 'run the council', 'council review', 'pressure-test this', 'stress-test this', 'war room this', or when facing a genuine decision with stakes and tradeoffs
Design conversational AI chatbots including intent recognition, slot filling, dialogue flow, and response generation. Use this skill when the user needs to build a chatbot, design conversation flows, implement intent classification, or improve chatbot accuracy — even if they say 'build a chatbot', 'our bot doesn't understand users', 'design a FAQ bot', or 'improve our chatbot's responses'.
Adversarially reviews whatever you bring it — prompts, strategies, plans, arguments, decisions, or workflows — and returns a structured, honest critique with scored dimensions and specific improvements. Combines a full prompt health check rubric with devil's advocate challenge modes for ideas, strategies, and decisions. Trigger this skill when the user says: "devil's advocate", "challenge this", "stress test", "poke holes in", "critique this", "what's wrong with", "find the flaws", "red team this", "what am I missing", "steelman the opposition", "evaluate my prompt", "review my prompt", "prompt health check", "rate my prompt", "is this a good prompt", "how can I improve this prompt", "evaluate this plan", "what could go wrong", "pre-mortem", "pressure test", "tear this apart", or any request to find weaknesses, gaps, or blind spots in something the user has shared. Also trigger when the user pastes a prompt, plan, strategy, or argument and asks for feedback — even without using these exact words.
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies. This skill orchestrates multiple AI agents (Codex, GLM, MiniMax, Amp, Grok) to work in parallel, maximizing throughput while avoiding conflicts. Covers task decomposition, agent selection based on capabilities, RALPH loop integration, conflict prevention, and real-world parallel execution examples.
Use when a complex development, fix, refactor, migration, verification, or OpenSpec change must be advanced across resumable checkpoints, recovered after context loss, or executed with durable progress and failure notes that stay inside the concrete OpenSpec change or archive directory. Also use when the user asks to generate a Claude Code `/goal`, Codex `/goal`, or do-long-task prompt. Do not use for short one-off edits, Q&A, simple formatting, changesets, release notes, or pure planning.
Use when the user wants two approaches raced head-to-head on a single shared metric — e.g. a classical/algorithmic lane vs an ML/learned lane, or any two strategies for the same task. Each lane runs its own analysis-first research loop confined to its lane, the lanes share a scoreboard and may borrow ideas across the boundary without abandoning their identity, and a shared eval keeps the head-to-head honest; loops until interrupted, reporting the current leader. Not for improving a single approach in isolation (use a single-track research loop), and not for picking between two finished artifacts in one shot (that is a one-time comparison).
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.
Comprehensive expertise for working with Microsoft's GenAIScript framework - a JavaScript/TypeScript-based system for building automatable LLM prompts and AI workflows. Use when creating, debugging, or optimizing GenAIScript scripts, implementing prompts-as-code, working with tools and agents, processing files (PDF, CSV, DOCX), defining schemas, or building AI automation workflows.
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.