A background service that continuously audits ClawHub skill download counts, detects ranking inflation, and flags skills with suspicious promotion patterns before users install them
ClawHub's download.increment mutation was exposed as an unauthenticated public RPC endpoint with no rate limiting. Silverfort's PoC pushed a malicious skill to #1 in its category with 3,900 executions across 50+ cities in 6 days. Even after the specific endpoint was patched, the fundamental problem remains: ClawHub has no integrity verification layer between the registry's popularity metrics and what users see. Download counts, star ratings, and trending positions can be gamed because no system validates that reported engagement represents real human installs. This service monitors the ClawHub registry continuously, builds statistical models of normal download velocity per skill category, and flags anomalies (sudden spikes, geographic concentration, bot-like install patterns) before they reach human users.
Demand Breakdown
Social Proof 1 sources
Gap Assessment
3 tools exist (SkillSieve, OpenClaw VirusTotal Integration, Bitdefender AI Skills Checker) but gaps remain: Analyzes skill CONTENT for malicious payloads. Does not monitor registry METADATA integrity (download counts, rankings, publisher reputation). A skill with clean code but fraudulently inflated downloads passes SkillSieve without issue.; Scans code artifacts only. Does not validate whether a skill's popularity metrics are authentic. A clean-code skill with inflated downloads and fake ranking position is invisible to VirusTotal..
Features3 agent-ready prompts
Competitive LandscapeFREE
| Product | Does | Missing |
|---|---|---|
| SkillSieve | Three-layer hierarchical triage framework that detects malicious AI agent skills using pattern matching, static analysis, and behavioral analysis. Achieves 0.800 F1 score. | Analyzes skill CONTENT for malicious payloads. Does not monitor registry METADATA integrity (download counts, rankings, publisher reputation). A skill with clean code but fraudulently inflated downloads passes SkillSieve without issue. |
| OpenClaw VirusTotal Integration | All published skills scanned using VirusTotal threat intelligence including Code Insight capability. Catches known malware signatures and suspicious code patterns. | Scans code artifacts only. Does not validate whether a skill's popularity metrics are authentic. A clean-code skill with inflated downloads and fake ranking position is invisible to VirusTotal. |
| Bitdefender AI Skills Checker | Consumer-facing skill safety checker from Bitdefender. Analyzes skills for security risks including data exfiltration, system modification, and obfuscation. | Content-focused scanner like SkillSieve. No registry integrity monitoring, no ranking fraud detection, no publisher provenance verification. |
Sign in to unlock full access.