The SkillMD blog
Practical, technical writing on building Agent Skills: the SKILL.md format, cross-agent installs, tooling, and where the registry is going.
Claude skills vs Cursor skills: what's actually different (2026)
Both read the same SKILL.md file. The differences are in where they load from, which frontmatter fields they honor, how you invoke them, and what happens in the cloud. A field-by-field comparison with a decision guide.
AI agent skills: the complete guide to what they are and how they work (2026)
A plain-English, technically exact explanation of AI agent skills: the SKILL.md format, progressive disclosure, how agents pick a skill, which tools support them, how they differ from prompts, rules, MCP, and plugins, and where to find safe ones.
Claude skills vs GitHub Copilot skills: directories, surfaces, and the instructions layer
GitHub Copilot reads Agent Skills from .github/skills, .claude/skills, and .agents/skills across the CLI, cloud coding agent, code review, VS Code, JetBrains, and Visual Studio. How that compares with Claude Code, and how skills relate to copilot-instructions.md, prompt files, and custom agents.
Claude skills vs Codex skills: paths, $ vs /, openai.yaml, and precedence
Claude Code and OpenAI Codex both run Agent Skills from the same SKILL.md, but they load from different directories, invoke with different syntax, and put agent-specific settings in different places. Every difference, with examples.
Cursor rules vs Cursor skills: when to use each (and how to migrate)
Rules are always-on .mdc files; skills are on-demand SKILL.md folders. What each one costs, how they trigger, what wins when they conflict, what /migrate-to-skills converts, and a rule-of-thumb for splitting your existing rules.
Agent skills vs MCP servers: knowledge vs tools, and when you need both
MCP gives an agent tools and data through a running server. A skill gives it a procedure through a text file. Cost, latency, security, portability, and a decision table for choosing, plus how the two compose in Claude Code, Cursor, and Codex.
Claude Code skills vs plugins vs subagents vs commands vs CLAUDE.md
Five ways to customize Claude Code, one decision rule. What each mechanism loads, when it loads, what it costs, and how skills now absorb commands and can fork into subagents. With a table for every combination.
One SKILL.md for every agent: the 2026 compatibility matrix
Which frontmatter fields Claude Code, Cursor, Codex, GitHub Copilot, Gemini CLI, and the rest actually honor, where each one reads skills from, and a portable template that loads unchanged in all of them.
Skill packs: shipping an outcome, not just a skill
What a pack is, why official and community packs exist side by side, and when to bundle several narrow skills instead of publishing one big one.
How search finds the right skill
Inside the registry's hybrid search: why typos, paraphrases, and vague requests still surface the right skill, and what that means for how you write a description.
What a verified badge actually means
The registry checks every skill with three separate auditors on one shared verdict scale. Here's what each one looks for and why lint alone isn't the safety review.
Anatomy of a SKILL.md file
A field guide to the SKILL.md format: the frontmatter fields that get parsed, how lint rules and quality scores work, and the mistakes that quietly tank a score.
Install one skill in every agent you use
Every coding agent keeps skills in its own folder. Here's how to install one with the skillmd CLI, target specific agents, or skip the CLI entirely.
Using SkillMD over MCP
Connect an agent to the SkillMD registry over MCP so it can search, inspect, and install skills on its own, no copy-pasting required.
Publishing a skill to the registry
How to take a skill from a local SKILL.md to a listed entry on the registry: login, dry-run, publish, and what happens after you submit.
Debugging skills: lint, scan, and the quality score
How to diagnose a failing or low-scoring skill locally with skillmd lint, rules, and scan before you publish, instead of guessing.
Build your first API-integration skill
A walkthrough for wrapping a public REST API in a SKILL.md file, from scaffolding to lint to publish, with a worked weather-API example.
Common SKILL.md mistakes and how to avoid them
The mistakes skillmd lint catches most often in SKILL.md files, why each one costs points, and the exact fix for each.
The case for an open skill format
As more coding agents ship, the industry needs a shared, plain-text way to define agent skills instead of rebuilding the same capability for every tool.
Cross-platform skills: Claude, Cursor, and beyond
SKILL.md is plain YAML plus Markdown, so one file works across agents. Where each agent stores skills, how to install and test across them.
Writing skills that get used
Principles for authoring a skill people actually install: narrow scope, a precise description, clear steps, explicit inputs and failure handling.