resources
guides, explainers, and deep dives on cross-vendor skill discovery, the SKILL.md ecosystem, and building solo with AI.
guides
all guides →the build-solo series: ship a real product with AI, one day at a time.
- day 1may 24, 2026
Day 1 — The Live URL First
The 5-step Day 1 flow from the @ImplexaAI build-solo series — Claude Code, Claude Pro, the one-sentence prompt, push + deploy, optional design mode. Get a live URL on the internet in 10 minutes without writing code.
read - day 2may 25, 2026
Day 2 — Frontend, Demystified
The 5-step Day 2 flow from the @ImplexaAI build-solo series. What frontend actually is. Next.js as the standard. Describing UI in plain English. Iterating by talking, not editing CSS. And the new gatekeeping skill: taste.
read - day 3may 26, 2026
Day 3 — Backend, Demystified
The 5-step Day 3 flow from the @ImplexaAI build-solo series. What backend actually is. Why you rent it in 2026 instead of building it. Supabase + Stripe + Resend = 200 lines of glue. The new gatekeeping skill: knowing what to compose.
read - day 4june 2, 2026
Day 4 — Git Is Your Time Machine
The 5-step Day 4 flow from the @ImplexaAI build-solo series. What Git actually is. Why it's a time machine, not a chore. Commits, branches, and undo — all in plain English, with Claude running the commands. The safety net that lets you build fearlessly.
read - day 5june 4, 2026
Day 5 — Deployment, the Magic Moment
The 5-step Day 5 flow from the @ImplexaAI build-solo series. Deploying is one click and sixty seconds. The real trap is the backend still on localhost. Fix it, add analytics, skip the domain, and text the live link to one real person — the moment you stop learning to build and just ship.
read - day 6june 5, 2026
Day 6 — Ship Day: End-to-End
The Day 6 capstone from the @ImplexaAI build-solo series. Ship one real full-stack app — a to-do list with login + database, deployed — and build the whole thing by prompting an AI. Next.js + Supabase + Vercel, three pieces, one live app. The day you stop learning to build and just ship.
read - day 17june 17, 2026
Day 17 — The Agent Did the Step I Forgot to Ask For
The Day 17 moment from the @ImplexaAI build-solo series. I let an agent migrate a friend's site off a vibe-coding platform onto my own stack — and without being asked, it set up Google Search Console and submitted the sitemap. The boring SEO last mile everyone forgets. This is the line between a tool that does the task and an agent that owns the outcome — and the exact public agent that ran it.
read - day 20june 22, 2026
Day 20 — Make Your AI Agent Get Better: The Feedback Loop
Most AI agents repeat the same mistake forever, because the correction lives in your head, not the agent. Here's the four-move feedback loop that turns a tool into a teammate: every run it shows its work, asks one question, you correct it once in plain English, and the note persists into the next run. From the @ImplexaAI build-solo series.
read - day 23june 27, 2026
Day 23 — Give Your AI Agent the Work, Not the Keys: The Irreversible-Action Approval Gate
An AI agent deleted someone's entire production database this week — backups and all — with one command nobody approved. It didn't go rogue; it did exactly what it was allowed to do. Here's the one gate that stops it from happening to you, the honest caveat that keeps the fix from becoming useless, and why credentials are the real blast radius. From the @ImplexaAI build-solo series.
read - day 25july 1, 2026
Day 25 — Your AI Isn't Lying, It's Guessing: How to Stop Your Agent Making Things Up
A giant firm just pulled an AI report after only five of its forty-five sources turned out to be real. Your own agent is doing a quieter version of the same thing — because it answers from a fuzzy memory of the internet instead of your real files. Here's the one plain-English rule (grounding / RAG) that stops most of it: point it at your real sources and let it say 'I don't know'. From the @ImplexaAI build-solo series.
read
blog
all posts →short, practical explainers on Claude skills, SKILL.md, and the workflow loop.
- blogjuly 10, 2026
Claude Code best practices: a practical 2026 guide
Claude Code best practices, from CLAUDE.md and plan mode to context management, verification loops, subagents, and hooks, so your sessions ship code you trust.
read - blogjuly 8, 2026
Claude Code GitHub Actions: automate PRs and CI
Claude Code GitHub Actions runs Claude in your CI to review PRs, fix issues, and answer @claude mentions. Here is how to set it up and the YAML you need.
read - blogjuly 6, 2026
What is the Claude Agent SDK, and do you need it?
The Claude Agent SDK lets you run Claude Code's agent loop inside your own app. What it is, how it differs from the Claude API, and when you actually need it.
read - blogjuly 5, 2026
SKILL.md vs CLAUDE.md: what each file does in Claude Code
SKILL.md vs CLAUDE.md in Claude Code: CLAUDE.md is always-on project memory, SKILL.md is an on-demand capability Claude loads only when a task matches it.
read - blogjuly 4, 2026
Claude skills vs plugins: which should you build?
Claude skills vs plugins, settled: what each one is, when a skill is enough, when to package it as a plugin, a decision table, and how the two fit together.
read - blogjuly 3, 2026
Claude Code settings.json: the complete config guide
Claude Code settings.json controls permissions, environment variables, the model, and hooks. Where each file lives, how they layer, and what the keys do.
read - blogjuly 2, 2026
Claude Code plan mode: how and when to use it
Claude Code plan mode is a read-only state where Claude proposes a plan and changes nothing until you approve. How to turn it on, and when it earns its place.
read - blogjuly 2, 2026
Claude Code memory: how CLAUDE.md and auto memory work
How Claude Code memory works: CLAUDE.md files, the load order and @import syntax, auto memory in ~/.claude, and when to use memory vs a skill, rule, or hook.
read - blogjuly 1, 2026
Claude skills vs subagents: which should you use?
Claude skills vs subagents, settled: what each one is, a decision table for your task, and how to run them together, with real SKILL.md and subagent examples.
read - blogjune 28, 2026
Claude Code slash commands: how to create one
Claude Code slash commands are Markdown files in .claude/commands you invoke with a slash. How to write one, pass arguments, and how skills changed them.
read - blogjune 26, 2026
How to add an MCP server to Claude Code
How to add an MCP server to Claude Code: the claude mcp add command for HTTP and stdio servers, the three config scopes, OAuth sign-in, and the safety check.
read - blogjune 25, 2026
Claude Code plugins: what they are and how to use them
Claude Code plugins bundle skills, commands, agents, hooks, and MCP servers into one unit. What they are, how to install one, and how to build your own.
read - blogjune 24, 2026
Claude Code hooks: automate your workflow reliably
Claude Code hooks run commands, HTTP calls, or prompts at fixed points in a session, so linting and guardrails fire every time. Here is how to set them up.
read - blogjune 23, 2026
Claude Code subagents: what they are and how to build one
Claude Code subagents are focused helpers that Claude spawns in their own context to handle one job. What they are, when to use them, and how to build one.
read - blogjune 23, 2026
Why AI Agents Still Need a Human in the Loop (2026 Data)
AI agents hit perfect benchmark scores without solving anything, and slip to a coin flip on long tasks. Here's the 2026 data — and why a human gate wins.
read - blogjune 22, 2026
How to schedule Claude Code to run on its own
How to schedule Claude Code to run on its own: desktop scheduled tasks, the loop command, cloud routines, and the limit each option quietly hides from you.
read - blogjune 21, 2026
AGENTS.md: the open standard for coding agent instructions
AGENTS.md is the open format that tells AI coding agents how to work in your repo. What it is, where it lives, what to put in it, and how it beats a README.
read - blogjune 20, 2026
How to build an AI agent with Claude (no code)
Build an AI agent with Claude, no code required: what an agent actually is, why Claude Code already gives you the loop, and how to make one run on its own.
read - blogjune 15, 2026
How to use Claude skills: trigger, invoke, and manage
How to use Claude skills: how Claude triggers the right one, invoking one yourself with a slash command, passing arguments, and fixing skills that never fire.
read - blogjune 12, 2026
Claude Code skills: how they work and how to build one
Claude Code skills are SKILL.md folders that load on demand in your terminal. Where they live, the three scopes, how to invoke them, and how to build one.
read - blogjune 9, 2026
Where to find Claude Skills: 7 places to look in 2026
Where to find Claude Skills in 2026: Anthropic's official library, the Claude Code marketplace, GitHub, third-party marketplaces, and cross-vendor search.
read - blogjune 8, 2026
Claude Skills vs MCP: which one should you actually build?
Claude Skills vs MCP, settled. What each one is, a decision table for your use case, and how to run them together, with real SKILL.md and MCP examples. (2026)
read - blogjune 3, 2026
Commoditized intelligence is the labs' moat, not their problem
The consensus says cheap tokens kill the frontier labs. I think the opposite: commoditized intelligence is the weapon Claude Code and Codex use to win.
read - blogjune 3, 2026
Claude Skills examples: 10 real SKILL.md files to copy
Claude Skills examples you can copy: 10 real SKILL.md files for PR review, sales prep, standups, changelogs, and invoices, each with the full instruction pack.
read - blogjune 2, 2026
How to use your Claude skills in Cursor, Codex, and Gemini
Your Claude skills already work in Cursor, Codex, and Gemini. Here are the SKILL.md portability rules, the directory map per agent, and how to skip the copy-paste.
read - blogjune 2, 2026
How Many Claude Skills Is Too Many? The Real Answer
How many Claude skills is too many? It's a context-budget problem, not a number. Here's the real cap, why skills stop triggering, and how to fix it for good.
read - blogmay 18, 2026
Claude Skills explained: the 2026 builder's guide
Claude Skills are reusable instruction packs loaded on demand. What's in a SKILL.md, skills vs prompts and MCP, where to find them, and how to build one.
read - blogmay 18, 2026
how to create a claude skill (step-by-step)
a 7-step walkthrough for creating your first Claude Skill, from install through recording, interview, activation, and sharing with your team.
read
deep dives
cornerstone pieces on why a cross-vendor skill graph exists and how it ranks.
- cornerstonemay 28, 2026
the two consolidation problems in claude code
The HN debate on Claude Code consolidation surfaced a real pain point. But it's actually two distinct problems wearing the same trenchcoat: where prompts live (architecture, which Anthropic is solving) and how you find good ones (discovery, which they aren't). This is about the second one.
read - cornerstonemay 27, 2026
SkillRank: how implexa scores 22,000+ skills across vendors
SkillRank is implexa's quality + relevance score for every AI skill in the cross-vendor graph. four signals, one number, ranked + searchable inside Claude Code and Codex.
read - cornerstonemay 25, 2026
the first ambient cross-vendor skill recommender, with privacy-by-discard
implexa is the ambient cross-vendor SKILL.md recommender. one index across 6 sources. inline-apply, no install. privacy by discard.
read - cornerstonemay 20, 2026
what is a skill graph? definition, examples, and how AI teams use them
a skill graph is a structured map of reusable workflows or capabilities that can be composed, shared, and measured. used in AI agents and workforce planning.
read