Most development teams don't use a single AI agent. One developer prefers Claude Code for agentic tasks, another uses Cursor for inline editing, a third uses Cline for VS Code. The result: skills written for one tool don't reach the others, and the team's AI procedures fragment.
MDSkill solves this. A single skill can be installed for multiple agents, and the underlying SKILL.md format works across all of them.
How the same skill works across agents
A SKILL.md is agent-agnostic. The format — Purpose, Instructions, Output format, Examples — is plain Markdown. The agent reads it and follows the procedure. It doesn't matter which agent reads it.
The difference is where the file lives. Each agent reads from a different path:
| Agent | Skills directory |
|---|---|
| Claude Code | .claude/skills/ |
| Cursor | .cursor/skills/ |
| Cline | .cline/skills/ |
| GitHub Copilot | .copilot/skills/ |
| Windsurf | .windsurf/skills/ |
The MDSkill CLI handles routing. When you install with the -a flag, it places the skill in the right directory:
# Install for Claude Code
npx mdskill add owner/repo/skill-name
# Install for Cursor
npx mdskill add owner/repo/skill-name -a cursor
# Install for Cline
npx mdskill add owner/repo/skill-name -a cline
Sharing skills across a team
The cleanest approach for multi-agent teams: commit all skills directories to your repository.
.claude/skills/
└── pr-security-reviewer/SKILL.md
.cursor/skills/
└── pr-security-reviewer/SKILL.md
.cline/skills/
└── pr-security-reviewer/SKILL.md
Every developer who clones the repo gets the same skills for whichever agent they use. When a skill is updated, one commit propagates it to everyone.
If you want to avoid duplicating files, symlink them:
# Link Cursor's version to Claude's
ln -s ../../.claude/skills/pr-security-reviewer .cursor/skills/pr-security-reviewer
Installing for multiple agents at once
To install the same skill for all agents your team uses, run the install command once per agent:
SKILL="owner/repo/skill-name"
npx mdskill add $SKILL
npx mdskill add $SKILL -a cursor
npx mdskill add $SKILL -a cline
Or script it into your project setup:
#!/bin/bash
# setup-skills.sh — run once after cloning
SKILLS=("owner/repo/security-auditor" "owner/repo/doc-generator")
AGENTS=("" "-a cursor" "-a cline")
for skill in "${SKILLS[@]}"; do
for agent_flag in "${AGENTS[@]}"; do
npx mdskill add $skill $agent_flag
done
done
When agent-specific skills make sense
Most skills work identically across agents. But some workflows are better matched to specific tools:
| Task | Best agent | Why |
|---|---|---|
| Multi-file refactor | Claude Code or Cline | Autonomous, reads entire codebase |
| Inline review as you type | Cursor | Sees the line you're editing |
| PR description generation | Any | All work well in chat mode |
| Architecture audit | Claude Code or Cline | Can navigate deeply and run commands |
For agent-specific tasks, you can write agent-variant skills — same base procedure, adapted instructions for what that agent can do.
Keeping skills in sync
When you update a skill (fix a bug in its procedure, improve the output format), update it in one place and redistribute. The MDSkill CLI handles this:
# Re-install updated skill for each agent
npx mdskill add owner/repo/skill-name
npx mdskill add owner/repo/skill-name -a cursor
Commit the updated SKILL.md files to propagate the change to your team.
What's next?
- Browse the skill directory to find skills compatible with your agents
- Install your first skill — two minutes from search to active
- Build a skill your whole team can use and publish it