AI for Developers
Agent skill (SKILL.md) Open source 2.7k

skill-scanner

A scanner for agent skills that looks for prompt injection and data exfiltration.

Install

For this project
$ npx skills add cisco-ai-defense/skill-scanner -a claude-code
For every project
$ npx skills add cisco-ai-defense/skill-scanner -a claude-code -g

Or copy the skill folder into .claude/skills/ (project) or ~/.claude/skills/ (all projects).

For this project
$ npx skills add cisco-ai-defense/skill-scanner -a codex
For every project
$ npx skills add cisco-ai-defense/skill-scanner -a codex -g

Or copy the skill folder into .agents/skills/ (project) or ~/.agents/skills/ (all projects).

For this project
$ npx skills add cisco-ai-defense/skill-scanner -a cursor
For every project
$ npx skills add cisco-ai-defense/skill-scanner -a cursor -g

Or copy the skill folder into .agents/skills/ (project) or ~/.cursor/skills/ (all projects).

When to use

Run it on third-party skills before you install them, or in CI for skills your team publishes.

Skill Scanner is Cisco AI Defense's security scanner for agent skills. It looks for prompt injection, data exfiltration and malicious code. It combines YAML and YARA-X pattern rules, AST and dataflow analysis, an optional LLM judge, and a bounded CEL decision layer that correlates findings before the LLM step.

It reads the Codex and Cursor skill formats, and with --lenient it also scans Claude Code commands and flat markdown skill repos. You can run it from the CLI, a Python SDK or an interactive wizard, and it has low-noise and quiet presets. Cisco is clear that it is best effort. A scan with no findings does not mean the skill is safe.

Related skills

All skills