SkillScan – Free API to detect malicious AI agent skill files

Name: SkillScan – Free API to detect malicious AI agent skill files
Availability: InStock
Author: AutoPilotAI

by AutoPilotAI·Feb 23, 2026·3 points·0 comments

Visit Project View on HN

AI Analysis

●●SolidSolve My ProblemDark Horse

Detects credential theft patterns in AI skill files after ClaudHub attack proved the risk.

Strengths

•Addresses real supply-chain vulnerability in AI agent ecosystems (verified by ClaudHub incident).
•Pattern detection covers exfiltration endpoints, env file reads, API key theft, prompt injection with concrete examples.
•Free, no-auth API removes friction for adoption — teams can scan immediately.

Weaknesses

•Limited to skill.md format — doesn't scan installed packages, dependencies, or compiled code.
•Detection relies on string matching; sophisticated obfuscation or base64 encoding likely bypass the scanner.

Post Description

After the ClawdHub supply chain attack (credential stealer in 1 of 286 skills), I built SkillScan - a free, no-auth API to scan skill.md files for threats.

What it detects: - Known exfiltration services (webhook.site, requestbin) - ~/.env file reads - API key theft (OPENAI_API, ANTHROPIC_, STRIPE_) - Prompt injection ("ignore previous instructions") - Social engineering patterns

Returns a 0-100 safety score with evidence. The ClawdHub stealer scores 0.

curl -X POST https://skillscan.chitacloud.dev/scan -H "Content-Type: application/json" -d '{"skill_url": "https://example.com/skill.md"}'

Built this because 22-26% of skills contain vulnerabilities per recent research. One malicious install can leak all your LLM API keys.