All Problems

Coding CLIs / Code Agents

The hottest category right now. Ten+ serious CLI agents competing across three tiers. SWE-bench Pro (standardized) is necessary but no longer sufficient — METR found ~50% of SWE-bench-passing PRs would NOT be merged by real maintainers. Rankings weight benchmarks alongside practical tests, adoption, safety, and independent evaluations.

Open full report →

Web Browsing / Browser Automation

The category has split into four lanes: full-autonomy agents (Browser Use, Skyvern), MCP/CLI tools for coding agents (Chrome DevTools MCP, Playwright MCP/CLI, Vercel Agent Browser), frameworks/SDKs for building products (Stagehand), and consumer agentic browsers (BrowserOS). CLI-over-MCP is settled consensus (13+ independent sources). Browser Use hits 1M+ weekly PyPI downloads — unchallenged in Lane 1. BrowserOS just crossed 10K stars as the Lane 4 leader.

Open full report →

Product / Business Development

Eight distinct lanes confirmed by independent traction data: Research/Extraction (Firecrawl #1, Exa #2), Enterprise Operating Surface (mcp-atlassian #1, Rovo #2), Startup Operating Surface (Notion), CRM (HubSpot #1, Salesforce #2, Dynamics 365 #3 NEW), Business Automation (Zapier #1, n8n watch), Product Analytics (PostHog #1, Amplitude #2 NEW, Mixpanel watch), Project/PM (Linear #1, Monday watch, Asana watch), and Communication (Slack, real but early).

Open full report →

Teams of Agents / Multi-Agent Orchestration

Five distinct segments with almost no cross-over: (1) Python agent frameworks — build multi-agent systems in code (LangGraph #1, OpenAI Agents SDK #2, Pydantic AI #3, CrewAI #4, plus cloud-native: Strands/AWS, ADK/GCP, Semantic Kernel/Azure); (2) TypeScript framework — Mastra (no competitor); (3) Autonomous coding agents — delegate software development to an agent (OpenHands, Factory AI); (4) Parallel agent IDEs — run multiple coding agents simultaneously (Emdash, ccpm, Superset); (5) Workflow automation — orchestrate integrations visually (n8n, Sim Studio). Ranking all on a single list is misleading — each serves a different buyer.

Open full report →

UX / UI

Five lanes: (A) read-only Figma context (Official #1, Framelink #2), (B) bidirectional write-access (Console MCP #1, Grab #2, figma-use #3), (C) alternative platforms (Penpot, Excalidraw), (D) specialized design-to-code agents (Kombai — 75–80% fidelity), (E) AI-native design creation (Google Stitch — Figma stock -8.8%, Onlook). Uber uSpec remains strongest enterprise validation. Google Stitch is provisional (2 days old).

Open full report →

Software Factories

Autonomous coding agents that plan, write, test, and ship code with minimal human oversight. Claude Code leads on benchmarks, community signal, and platform distribution (Apple Xcode). Cursor leads on revenue and event-driven automation. Gemini CLI is the free-tier disruptor. The category has split into CLI-first (Claude Code, Codex CLI, Gemini CLI), IDE-integrated (Cursor, Copilot, Cline), open-source (OpenHands), and enterprise-managed (Augment, Factory). SWE-bench Verified is dead — Pro is the new standard.

Open full report →

Search & News

Web search, scraping, and deep research tools for AI agents. The category has split into three lanes: search APIs (Brave, Exa, Tavily), scrape/crawl tools (Firecrawl, Crawl4AI), and deep research APIs (Parallel, Perplexity Sonar). Most serious agent workflows need tools from the first two lanes. MCP support is table stakes — the real differentiators are benchmark quality, latency, index independence, and license. Deep research lane is still immature — WideSearch academic benchmark shows near 0% success on broad tasks.

Open full report →

Marketing

Skills for SEO, content optimization, ad copy, social media calendars, competitor analysis, and growth automation.

Open full report →

Business

Skills for pitch decks, financial modeling, contract review, OKR frameworks, invoicing, and business operations.

Open full report →

Content & Writing

Skills for blog posts, newsletters, technical writing, style guide enforcement, and editorial workflows.

Open full report →

Research

Deep research agents, academic tools, and research infrastructure. The category has split: platform deep research (Perplexity, OpenAI, Google), open-source agents (GPT Researcher, Tongyi, STORM), academic specialists (Elicit, Consensus), and infrastructure (Tavily, Firecrawl). Speed, citation quality, and self-hosting are the real differentiators.

Open full report →

Automation

Three sub-categories with distinct buyers: Workflow Automation (n8n #1, Activepieces, Zapier), Code-First Orchestration (Windmill, Trigger.dev, Inngest, Kestra), Agent Integration (Composio, Pipedream MCP). n8n dominates overall with 180K stars, n8n-mcp (15.4K stars), and dedicated Claude Code skills. Inngest leads npm downloads (499K/wk). Composio is provisional — zero HN traction despite 27K stars.

Open full report →

Security

Skills for SAST scanning, secret detection, agent/MCP security scanning, and offensive security. The category splits into four sub-themes: SAST/code scanning (Semgrep MCP #1), secret detection (GitGuardian MCP #1), agent/MCP security scanning (Snyk Agent Scan #1), and offensive security (HexStrike AI #1). Agent security scanning is the fastest-growing sub-theme — these tools scan your agents, skills, and MCP servers, not your application code.

Open full report →

Documentation

Docs-as-code frameworks, API documentation generators, documentation SaaS platforms, and documentation automation tools. The category splits into four overlapping lanes: OSS docs frameworks (Fumadocs #1, Starlight #2, Docusaurus #3), API docs (Fern #4, Redocly #6, Swagger UI #8), SaaS platforms (Mintlify #5, GitBook #10), and automation (Promptless #9). Fumadocs is the momentum winner — 3x YoY growth and 5-open-issues maintenance make it the clear pick for Next.js teams.

Open full report →

Data & Analytics

AI-powered data analysis tools, reactive notebooks, BI-as-code platforms, conversational data agents, and ML training aids. The category has split into reactive notebooks (Marimo), AI visualization (Data Formulator), BI-as-code (Evidence, Observable), app deployment (Streamlit), conversational data (PandasAI), and prompt-to-ML (Plexe). Marimo is the clear #1 — strongest combined signal across stars, downloads, HN attention, and independent validation.

Open full report →

Personal Assistants

How do I interact with AI for everyday tasks? ChatGPT, Claude, Gemini, and emerging contenders like OpenClaw.

Open full report →

Memory Systems

How do agents remember context across sessions? Vector DBs, context management, and persistent memory for AI workflows.

Open full report →

Performance

How do I profile, benchmark, and optimize AI/agent workloads? Speed and efficiency tools for production deployments.

Open full report →

Analytics & LLM Tracing

How do I observe and trace LLM calls and agent runs? PostHog, Braintrust, LangSmith, Helicone, and more.

Open full report →

Web Development & UI Frameworks

How do I build AI-powered UIs? Frontend frameworks and tools for shipping AI products — Vercel AI SDK, Streamlit, v0, Bolt, Lovable.

Open full report →

Agent Harnesses

How do I orchestrate and run agents? Frameworks for building, deploying, and managing AI agent workflows — LangChain, CrewAI, Pydantic AI, Claude Agent SDK.

Open full report →

Knowledge Management

How do I organize and retrieve team knowledge? Notion, Google Workspace, Obsidian — the foundations MCP tools build on.

Open full report →

AI Adoption & Best Practices

How do I adopt AI effectively? Meta-tracking, best practices, ecosystem navigation. SkillPack itself lives in this problem space.

Open full report →