Every claim and every figure the video puts on screen, with a primary source for each. Checked 8 August 2026.
A Skill is a folder. Inside it, a required SKILL.md file carries metadata and
instructions, and alongside it a Skill may bundle a scripts/ directory of executable
code, a references/ directory of documentation, and an assets/ directory of templates
and supporting files.
“Agent Skills are modular capabilities that extend Claude’s functionality. Each Skill packages instructions, metadata, and optional resources (scripts, templates) that Claude uses automatically when relevant.”
Source: Agent Skills overview, Claude Platform Docs
Two required frontmatter fields, name and description, are what Claude matches a
request against. The description has to say both what the Skill does and when to use it.
Source: Agent Skills overview — Skill structure
Skills load in stages, which the documentation calls progressive disclosure. Only the
name and description sit in context at startup, at roughly 100 tokens per Skill. The body
of SKILL.md loads only when the Skill is triggered. Bundled reference files and scripts
load only when they are actually read or run, and a script’s code never enters the context
window at all, only its output.
Source: Agent Skills overview — How Skills work
Anthropic publishes a public repository of Agent Skills. At the time of checking, the
skills/ directory contains seventeen Skills:
algorithmic-art, brand-guidelines, canvas-design, claude-api, doc-coauthoring,
docx, frontend-design, internal-comms, mcp-builder, pdf, pptx, skill-creator,
slack-gif-creator, theme-factory, web-artifacts-builder, webapp-testing, xlsx.
Source: anthropics/skills, skills/ directory
Every Skill named in this video appears in that list. The spreadsheet Skill is xlsx, the
presentation Skill is pptx, and web app testing is webapp-testing.
The four document Skills — PowerPoint, Excel, Word and PDF — are also available pre-built
on claude.ai and through the Claude API, where they are referenced by the skill ids pptx,
xlsx, docx and pdf.
Source: Agent Skills overview — Available Skills
Every chapter that names a Skill puts its repository path and its install command on screen. Those come from the repository’s own marketplace definition rather than from the directory listing, and the difference matters: the marketplace publishes three plugins, each carrying a set of Skills, so a Skill is installed by its set and not by its own name.
In Claude Code:
/plugin marketplace add anthropics/skills
/plugin install document-skills@anthropic-agent-skills
/plugin install example-skills@anthropic-agent-skills
| Plugin | Skills it carries |
|---|---|
document-skills |
xlsx, docx, pptx, pdf |
example-skills |
algorithmic-art, brand-guidelines, canvas-design, doc-coauthoring, frontend-design, internal-comms, mcp-builder, skill-creator, slack-gif-creator, theme-factory, web-artifacts-builder, webapp-testing |
claude-api |
claude-api |
Source: anthropics/skills, .claude-plugin/marketplace.json
and anthropics/skills, README
Two things worth knowing alongside that, because they mean some of these need no install at all:
claude-api Skill already ships bundled with Claude Code.
Source: Claude API skillpptx, xlsx, docx and pdf.
Source: Agent Skills overviewA custom Skill of your own is a directory rather than an install: ~/.claude/skills/ for
personal use, or .claude/skills/ inside the project it belongs to.
Source: Extend Claude with skills, Claude Code docs
MCP is the Model Context Protocol.
“MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems. Using MCP, AI applications like Claude or ChatGPT can connect to data sources (e.g. local files, databases), tools (e.g. search engines, calculators) and workflows (e.g. specialized prompts) — enabling them to access key information and perform tasks.”
Source: What is the Model Context Protocol?, modelcontextprotocol.io
The mcp-builder Skill in the official repository is the one the video refers to for
creating those integrations.
Source: anthropics/skills, skills/ directory
The claude-api Skill supplies current API reference material, SDK documentation and best
practices, and it covers eight languages: Python, TypeScript, C#, Go, Java, PHP, Ruby and
cURL. It detects a project’s language by examining project files, for example
requirements.txt for Python, tsconfig.json for TypeScript and go.mod for Go, and
loads only the matching documentation.
It ships bundled with Claude Code and is also installable from the open source repository.
Source: Claude API skill, Claude Platform Docs
The skill-creator Skill does more than write a SKILL.md. Its documented workflow runs
test cases against the Skill, grades the results, and iterates:
Source: anthropics/skills, skill-creator/SKILL.md
That is the basis for the point that the hard part is not writing the instructions, it is making them fire consistently.
The framing the video closes the argument on is Anthropic’s own:
“Skills are reusable, filesystem-based resources that give Claude domain-specific expertise: workflows, context, and best practices that turn a general-purpose agent into a specialist. Unlike prompts (conversation-level instructions for one-off tasks), Skills load on demand, so you don’t have to repeat the same guidance across conversations.”
Source: Agent Skills overview — Why use Skills
Skills are documented as one customisation surface among several, and the Claude Code
skills documentation cross-references the others directly: CLAUDE.md files, hooks,
subagents, MCP servers and plugins. A skill folder with a .claude-plugin/plugin.json
loads as a plugin and can bundle agents, hooks and MCP servers itself.
Skills live in ~/.claude/skills/ for personal use or .claude/skills/ for a project, so
a Skill can sit inside the repository it applies to.
Source: Extend Claude with skills, Claude Code docs
The video uses a handful of everyday examples to make a point concrete. They are the narration’s own hypotheticals, not measurements, and each is marked on screen as illustrative:
Nothing in the video depends on those numbers being exact.