Overview
Aguara Watch continuously scans 28,000+ AI agent skills across 5 public registries to track the real-world threat landscape for AI agents. All scans are powered by Aguara.Visit Aguara Watch
Explore the real-time threat dashboard for AI agent skills
What It Does
Aguara Watch provides continuous security monitoring of the AI agent ecosystem by:Scanning Public Registries
Monitors skills and configurations from major AI agent platforms:- Claude Skills — Community-contributed skills for Claude Desktop
- Cursor Extensions — Tools and workflows for Cursor IDE
- VS Code MCP Servers — Model Context Protocol servers for VS Code
- Windsurf Plugins — Plugins for Windsurf editor
- Community Registries — Open-source skill collections
Real-Time Threat Detection
Scans all content using the full Aguara detection engine:- 177+ detection rules across 13 categories
- 4-layer analysis — pattern matching, NLP, taint tracking, and rug-pull detection
- Confidence scoring for each finding (0.0-1.0)
- Severity classification from INFO to CRITICAL
Tracking Changes Over Time
Monitors skills for behavioral changes that could indicate supply-chain attacks:- Rug-pull detection — Skills that change to add malicious behavior
- Version tracking — Historical view of skill modifications
- Trend analysis — Emerging threat patterns across the ecosystem
Use Cases
Before Installing Skills
Check if a skill has known security issues before adding it to your configuration:- Visit watch.aguarascan.com
- Search for the skill name or author
- Review findings and confidence scores
- Make an informed decision
Security Research
Understand the current threat landscape:- What types of security issues are most common?
- Which registries have the highest risk profiles?
- How are attack patterns evolving over time?
Registry Maintenance
If you maintain a skill registry:- Identify skills that need security review
- Track improvements in skill security over time
- Set security standards for your registry
Key Metrics
28,000+ Skills
Comprehensive coverage across 5 major registries
Continuous Scanning
Regular re-scans to catch new threats and changes
177+ Rules
Full Aguara detection engine with all categories
Public Access
Free threat intelligence for the community
Threat Categories Monitored
Aguara Watch tracks all security categories supported by Aguara:| Category | Common Threats |
|---|---|
| Prompt Injection | Instruction overrides, role switching, jailbreaks |
| Credential Leak | API keys, private keys, database strings |
| Data Exfiltration | Webhook exfil, DNS tunneling, sensitive file reads |
| Supply Chain | Download-and-execute, reverse shells, privilege escalation |
| External Download | Binary downloads, curl-pipe-shell, auto-installs |
| MCP Attack | Tool injection, name shadowing, capability escalation |
| Command Execution | shell=True, eval, subprocess, PowerShell |
| SSRF & Cloud | Cloud metadata access, IMDS, internal IPs |
| Unicode Attack | RTL override, homoglyphs, zero-width sequences |
How It Works
- Collection — Automated crawlers fetch skills from public registries
- Scanning — Each skill runs through the full Aguara analysis pipeline
- Storage — Findings are stored with confidence scores and context
- Presentation — Web dashboard makes findings searchable and filterable
- Monitoring — Re-scans detect changes and rug-pull attempts
Powered by Aguara
Aguara Watch uses the same scanner you can run locally:- Same detection rules and analysis engines
- Same accuracy and confidence scoring
- Same offline, deterministic approach
- Same open-source transparency
Contributing Data
If you maintain a public skill registry and want to be included in Aguara Watch:- Ensure your registry has public access (HTTP/S or git)
- Open an issue at github.com/garagon/aguara
- Include registry URL and format (JSON API, git repo, etc.)
- Aguara Watch will add your registry to the scan rotation
Related
- Aguara MCP — Give agents security scanning capabilities
- CLI Reference — Scan your own skills locally
- Detection Rules — See what Aguara Watch checks for