Topic index
55 topics across 13 clusters
Curated reference pages on Claude Code, Cursor, MCP, RAG, agentic engineering, AI evals, and API infrastructure — each backed by installable rulesets and real GitHub source attribution.
Claude Code skills, plugins, hooks
SKILL.md across tools: what is real, what is portable, what still needs glue
The emerging SKILL.md standard is real, but native support and install paths still vary. Here is the verified cross-tool picture and how to author portable skills.
AGENTS.md: The Cross-Tool Standard Explained
What AGENTS.md is, which tools read it, how it compares to CLAUDE.md, and the symlink trick to make both work in the same project. Adopted by Codex CLI, Cursor, Gemini CLI, Aider, Windsurf, and more.
Autoresearch in Claude Code: The Complete Guide (2026)
Autoresearch is autonomous deep-research where Claude Code investigates a question for hours on its own. Here are the 4 production ports, how each works, and which to install.
The 30 Claude Code Skills Worth Installing (We Audited 100)
A real audit, not a listicle. We installed 100 community skills, ran them against 4 repos, and graded each on trigger precision, output quality, and security. Here are the 30 that survived.
The Claude Code Hooks Cookbook: 12 Copy-Paste Recipes
Twelve production-ready hooks for PreToolUse, PostToolUse, Stop, and SubagentStop events. Each recipe: when to use it, what it does, the JSON snippet to drop into settings.json.
The CLAUDE.md Playbook
What CLAUDE.md is, where it lives, how to write one that actually works, and stack-specific templates for Next.js, Python, Rails, and Go. Based on HumanLayer's <60-line standard and Karpathy's 109k-star example.
Every Claude Plugin Marketplace, Ranked (2026)
Eight Claude Code plugin and skill marketplaces exist in May 2026. Each has a different trust model, distribution mechanism, and creator-economy posture. This is the comparison nobody else has written.
How to Write a Claude Code Skill Description That Actually Triggers
The SKILL.md description field is Claude's only trigger signal. This is the engineering discipline behind descriptions that get invoked when they should, and don't get invoked when they shouldn't.
Your Claude Skill Isn't Triggering. Here Are the 8 Reasons.
Eight specific causes — symptom, diagnosis, fix — for the most common Claude Code skill failure mode. Based on Obra's December 2025 debugging post and confirmed against GitHub issue #11266.
The 10-Point Skill Security Checklist
Ten things to check before installing any community Claude Code skill. Based on Snyk's ToxicSkills finding (13.4% of public skills have a critical issue) and the Rules File Backdoor CVE class.
MCP servers
The 18 MCP Servers We'd Trust in 2026
Glama lists 23,451 MCP servers. Snyk found security issues in 66% of them. Here are the 18 we'd actually install — and the dozens we'd uninstall.
GitHub MCP Server: Official vs Alternatives
Anthropic archived their reference GitHub MCP server in 2025. github/github-mcp-server is the official replacement, with proper PAT scoping and OAuth flows. This is the comparison against gitmcp.io, cyanheads/git-mcp-server, and what each is for.
The Postgres MCP Server (and Why Anthropic's Was Archived)
Anthropic archived their reference Postgres MCP server in 2025 with a known SQL-injection bypass still in the wild. crystaldba/postgres-mcp is the replacement most teams ship. This is the comparison, the install, and the security model.
Are MCP Servers Safe? 66% Scan Findings and 30+ CVEs Later
Snyk found security issues in 66% of MCP servers they scanned. Anthropic and Ox Security publicly disagreed about whose problem it is. Here's the honest map.
MCP Servers: What They Are (in One Sentence) and Which to Install
An MCP server is a process that exposes tools, resources, and prompts to an LLM client over a defined wire protocol. Here's what to install — and what to skip.
The 'Pick 3 MCP Servers' Rule (and Why 3 Is the Magic Number)
A single MCP server burns 14,214 tokens before user input. Stack 5 of them and you're past 66,000 — a third of Claude Sonnet 4.5's context window. Here's the math and the fix.
Paid MCP Servers: The New Category That Doesn't Have a Directory Yet
Under 5% of MCP servers are monetized. No canonical directory of paid ones exists. MCPize takes 15%, AgenticMarket promises 10%, xpay 5%. Here's the live category map.
Cursor rules
Cursor Rules for Flutter 3.27 + Riverpod (2026)
Flutter 3.27 + Riverpod 2 + Dart 3.6 changed enough that pre-2025 .cursorrules ship code your linter rejects. Here is the AsyncValue + codegen + freezed rule, with diffs against what Cursor produces without it.
Cursor Rules for Next.js 15 + React Server Components (2026)
Next.js 15 made async APIs, dynamic IO, and React Compiler the default — and most Cursor rules still target Pages Router. Here's the App Router-only rule that catches RSC mistakes before they ship.
Cursor Rules for React + Vite + TanStack (Non-Next.js, 2026)
An HN comment from March 2025 surfaced a real gap: every Cursor rule for React assumes Next.js. Here is the rule for the React + Vite + TanStack Router/Query stack that doesn't.
Cursor Rules for Rust + Axum + sqlx (2026)
Most Cursor rules for Rust are 8 lines and say `idiomatic Rust`. Here is the Axum + sqlx + tower production rule that catches `.unwrap()` in handlers, missing error mappings, and N+1 queries before the build.
Cursor Rules for TypeScript 5.6: The Modern-Stack Rule Set (2026)
TypeScript 5.6 changed enough that pre-2025 .cursorrules files now produce code your linter rejects. Here are the strict-mode, satisfies-operator, branded-types rules that actually work, with before/after diffs.
RAG, vector DBs, embeddings
The 12 Embedding Models Worth Running (MTEB-Graded, 2026)
Voyage-3-large beats OpenAI text-embedding-3-large by 10.58% at 1/24 the storage. BGE-M3 is the self-host cost leader. Here are the 12 models worth picking from, with the numbers.
RAG Chunking Strategy 2026 (Recursive Beats Semantic)
The 2026 chunking benchmark from Vecta measured 69% retrieval accuracy for recursive 512-token chunking and 54% for semantic chunking. The default 'advanced' advice underperformed the baseline. This is the chunking guide that updates with the data.
RAG Inside a Claude Code Project (as a Skill)
Anthropic owns 'contextual retrieval' definitionally. RuleSell owns RAG-as-a-Claude-Code-skill pragmatically. Here is the one-click skill bundle pattern, with the indexing job and the install.
RAG Framework Picker: LangChain, LlamaIndex, DSPy, Haystack, RAGFlow
Five RAG frameworks dominate production usage in 2026. Each is good at something specific and bad at something else. This is the picker — what each framework optimizes for, when it loses, and the configuration we'd actually ship.
RAG-as-MCP-Server: The Missing Pattern (2026)
Exposing a retrieval pipeline as an MCP server is the cleanest way to share RAG across Claude Code, Cursor, Codex, and any other MCP-capable agent. Here is the pattern, with code.
Vector Database Picker: 12 Options + Decision Tree (2026)
pgvector beats Qdrant at 1M scale in Supabase's HNSW benchmark. 70% of teams should never leave Postgres. Here is the honest decision tree for the other 30%.
AI coding agents
aider CONVENTIONS.md Per-Stack Templates
aider reads CONVENTIONS.md at repo root before each chat. The file is the single best lever for getting consistent output. This is the per-stack template library — Next.js, Rails, Django, Go, Rust — plus the architect/editor mode pairing for each.
How to run Codex and Claude Code on the same project
A practical setup for using OpenAI Codex and Claude Code together: shared instructions, clean handoffs, and when each tool earns the keyboard.
The Best AI Coding Agents for Ollama and Local Models (2026)
Running an AI coding agent against a local LLM via Ollama is finally viable in 2026. This is the picker — aider, Cline, Continue, RooCode — with which models actually work, hardware requirements, and the cost math vs cloud APIs.
Agentic engineering, autoresearch
Agent harness engineering matters more than model shopping
If your agent gets context badly, uses tools badly, and never verifies its own work, the model upgrade will not save you. The harness is where reliability comes from.
Agentic Engineering: The Post-Vibe-Coding Paradigm (2026)
Karpathy called vibe coding 'passé' in April 2026 at Sequoia. The SERPs are still 6+ weeks behind. Here is what agentic engineering means, what tools it implies, and how to upgrade your workflow.
Creator monetization playbook for rules, skills, and MCP servers
If you want to make money from agent configs, treat them like maintained software products: claim authorship, ship a free tier, price the premium layer tightly, and prepare your payout stack early.
Why a Multi-Tool AI Config Marketplace Wins in 2026
Developers switch between Claude Code, Cursor, Cline, Aider, Codex, Gemini CLI, Windsurf, Continue, and 8 more tools. Single-tool hubs lose them. A multi-tool marketplace doesn't.
RuleSell vs free directories: where the catalog ends and the marketplace begins
Cursor Directory, awesome-cursorrules, ClawHub, and Smithery are all useful. RuleSell's bet is different: multi-tool discovery, quality scoring, and creator monetization.
Obra Superpowers Explained: What It Actually Does
Obra Superpowers is a Claude Code methodology framework with 120k+ stars on GitHub. It bundles subagent-driven development, TDD discipline, brainstorming, and verification skills into one installable package. This is what each piece does and when to use it.
YOLO Mode (--dangerously-skip-permissions): The Safe Playbook
Running Claude Code without permission prompts is the most productive setting and the most dangerous one. Here is when to use it, when not to, and the guardrails that make it safe.
LLM eval and observability
Evaluating Claude Code skills: trigger precision and output quality (2026)
A Claude Code skill has two failure modes: it never fires when it should, or it fires and produces generic output. Most teams test only one. Here's the eval bundle that covers both.
LLM-as-Judge Prompt Library (Per Use Case)
An LLM judge scores another LLM's output against a rubric. The theory is settled — pairwise beats Likert, binary beats fine-grained, judges drift. The tactical part — judge prompts you can copy-paste per use case — has not been published in one place. Until now.
LLM evals: the Hamel process encoded as rulesets (2026)
Hamel Husain's eval process: 60-80% of dev time on error analysis, custom annotation tools, binary judges, review 100 traces. Here's how to encode that as a tool-agnostic ruleset that survives the next acquisition.
Promptfoo alternatives after the OpenAI acquisition (2026)
OpenAI acquired Promptfoo in March 2026. ClickHouse acquired Langfuse in January. Two of the three biggest OSS eval tools changed hands in 8 weeks. Here's what to use now.
AI API infrastructure
Vercel AI SDK v4 → v5/v6 migration: the codemod-gap ruleset (2026)
The official codemod handles renames. It cannot migrate persisted DB messages, stopWhen logic, or the UIMessage parts array. Teams with persisted message history face 2-6 weeks of engineering work. Here's the gap map.
Anthropic prompt caching with the AI SDK: the 95%-off stack (2026)
Batch + cache stacks to 95% off on Anthropic system tokens — $1.00 → $0.50 → $0.05 per MTok. The math is in the docs but nobody publishes it. Here's the provider-wrapper, the dashboard recipe, and the cases where caching never hits.
LLM Gateway Decision Tree: OpenRouter, Vercel AI Gateway, Portkey, LiteLLM
Four LLM gateways dominate production usage in 2026. Each optimizes for a different problem and has a different pricing posture. This is the decision tree — when each one wins, what BYOK actually costs, and what to migrate when you outgrow your first choice.
OpenAI Assistants → Responses migration: the 2026-08-26 sunset (2026)
OpenAI's Assistants API sunsets August 26, 2026. There is no automated Threads → Conversations migration. Teams with tenant-per-assistant architectures face 2-6 weeks of engineering work. Here's the migration map.