AI Skill Scanner

AI Security SaaS

AI Skill Scanner audits every skill in ~/.claude/skills before it reaches your team.

Paste a skill repository URL or upload your local Claude skills folder. We inspect each file for malicious hooks, hidden exfiltration paths, hardcoded credentials, and suspicious subagent patterns.

View scanner workspaceOpen security dashboard

Why teams buy

  • Pre-installation policy check for third-party AI skill packs.
  • File-level evidence and risk scoring usable in security reviews.
  • Helps procurement and platform teams standardize AI tool due diligence.

Built for teams adopting AI coding workflows

agentseal proves demand as a CLI wedge. This hosted scanner helps orgs move from ad-hoc checks to repeatable governance.

The problem

  • Skills can execute shell commands, spawn subagents, and silently pull remote scripts.
  • A single risky skill can leak secrets from ~/.ssh, .env files, or private code during agent runs.
  • Manual review breaks down when teams evaluate dozens of third-party skill packs each month.

How the scanner works

  1. 01Ingests skill files from upload or GitHub repo URL.
  2. 02Detects exfiltration commands, secret exposure, and suspicious automation hooks.
  3. 03Produces a weighted risk score with actionable remediation guidance.
  4. 04Generates AI triage notes to accelerate security review handoffs.

Pricing

Simple plan for engineering teams running AI tool due diligence.

$19per month
  • Unlimited scans for one engineering team workspace.
  • Risk scoring, evidence trace, and remediation checklist.
  • Lemon Squeezy billing with instant browser unlock after payment.

FAQ

What exactly gets scanned?

We parse SKILL.md files, scripts, prompts, and config in each skill package. The scanner flags suspicious command chains, data exfiltration paths, secret patterns, and high-risk subagent orchestration.

Can we scan private skill repos?

Yes. You can upload your local ~/.claude folder directly from your machine, or paste a GitHub repository URL for remote analysis.

How fast is a full audit?

Typical scans complete in under 30 seconds for most skill packs. Reports include risk score, concrete evidence, and prioritized remediation steps.

Who is this for?

Engineering teams with 10-50 developers that are integrating AI coding tools and need a repeatable due-diligence process.