v0.8
Agent skills are powerful. And untrusted.
Run any agent skill. Safely.
Without installing it.
One command — fetch, scan, run, auto-clean.
$ skillx run github:skillx-run/skillx/examples/skills/name-poem "Your Name"
→ Resolving source...
✓ Resolved: name-poem
→ Scanning for security issues...
✓ Scan Result: PASS (30 rules checked)
✓ Using agent: Claude Code
✓ Injected 1 files to .claude/skills/name-poem
→ Launching agent...
[Claude Code running...]
✓ Agent completed.
✓ Cleanup done.
Why skillx?
No Install Needed
ephemeralRun any skill temporarily. It's fetched, injected, and auto-cleaned when your session ends. Nothing is permanently added to your project.
Security First
30 rulesEvery skill is scanned across 5 risk levels before injection. 30 rules detect prompt injection, credential theft, obfuscation, symlink escapes, and self-replication — dangerous patterns are blocked before they touch your system.
One Command
zero configNo setup, no config files. One command pulls from any Git host, scans, runs, and cleans up. That's it.
How skillx works
↓
Fetch Pull from any Git host or skill directory
⊘
Scan 23 security rules check before use
→
Inject Copy skill files into agent config
▶
Run Launch your agent with the skill
✓
Clean Auto-remove when session ends
The Agent Skill Ecosystem
Works with every agent. Pulls from every source.
Agents 32+
AI coding assistants, IDEs, and CLI tools
Sources 18+
Git hosts, registries, and skill directories
One CLI. Every agent. Every source.