position
Current position
Choose by capability boundary, not by format popularity.
claim
Use a Skill
Use an Agent Skill for reusable instructions, examples, references, assets, and scripts that should load only when a workflow is relevant.
evidence
Filesystem contract
The Agent Skills specification defines a SKILL.md package with optional scripts, references, assets, and progressive disclosure.
↗ Agent Skills
evidence
Client adoption
GitHub documents the open Agent Skills standard across Copilot cloud agent, code review, CLI, app, and agent modes in multiple IDEs.
↗ GitHub Docs
evidence
Distribution still moving
The current well-known discovery proposal is a draft whose v0.2 format intentionally breaks with v0.1 and adds artifact digests and archive safety guidance.
↗ Cloudflare
claim
Use MCP
Use MCP when a host must discover and call live tools, read dynamic resources, or use prompts across a local or remote client-server boundary.
evidence
Protocol boundary
The current MCP architecture defines JSON-RPC discovery plus tools, resources, and prompts over stdio or Streamable HTTP transports.
↗ Model Context Protocol
claim
Use both
Pair them when a portable Skill should teach the workflow while MCP supplies live data or actions.
evidence
Runtime counterpressure
MCP's official security guidance covers authorization-specific attack classes and mitigations that a remote capability must operate correctly.
↗ Model Context Protocol
claim
Qualification
A Skill is not automatically static or safe: it may include executable scripts, and support for pre-approved tools varies by agent implementation.
unknown
Missing evidence
No committed cross-client benchmark yet measures whether one non-trivial Skill and one MCP integration behave consistently across major coding agents.