Skills Security Scanner
Check any AI Skill for security risks before installing. Free, static, instant.
AI Skills unlock powerful capabilities — and can hide serious risks. Every AI skill deserves a security check before it touches your machine. Paste a SKILL.md, upload a Skill file or zip, or drop a GitHub link. Our static skills security scanner inspects source as plain text so you can decide with evidence, not guesswork.
Scan a SkillAdd a SKILL.md source to begin.Static only
Static onlySource code is read as plain text — nothing is executed.
Security Report
A detailed skills security report, not a pass/fail sticker.
Every scan produces a comprehensive report with the fields you need to make an informed install decision.
Score and recommendationGet a 0–100 risk score with a clear recommendation — from clean signals to avoid-until-reviewed.
Findings with evidenceEach detected risk shows category, severity, confidence, source location, and a recommended action.
Coverage boundariesSee exactly which files were scanned, which were skipped, and which analysis channels ran.
Review and shareCopy the report link, export JSON, generate a share image, contact support if a result looks wrong, or scan again anytime.
63
Decision
Review before installing
Issues were found that could impact safety or lead to unintended behavior.
Highest riskHardcoded credential
Potential secret exposed in uploaded Skill files
SourceUploaded Skill archive
Scope4 / 5 files
Scan modeArchive workspace scan
ExpiresSep 16, 2026
Coverage limited. 1 file was skipped and 1 reference was unresolved.
Important findings2
HighHardcoded credentialPotential secret exposed in uploaded Skill files
const token = "ghp_************"Highauth.ts:12MediumUnsafe command executionCommand built from user input
exec(`curl ${userInput}`)Mediumfetch.js:27Review full reportOpen the complete evidence report
Share reportCopy a link or export JSON
Risk Categories
Every skills security risk we check, listed here. No black boxes.
The skills scanner checks for the risk patterns that matter most when installing third-party Skills — each category independently weighted and scored.
Our detection draws on established security frameworks: OWASP Top 10 for LLM Applications (prompt injection), MITRE ATT&CK (credential harvesting T1552, data exfiltration T1041), and CWE-94 (code injection).
Prompt Injection
Hidden instructions that override model behavior, hijack system prompts, or manipulate agent outputs.
Weight1.00
Credential Harvesting
Patterns that steal API keys, environment variables, access tokens, or local secrets from your machine.
Weight1.25
Data Exfiltration
Network calls that send files, conversation history, or private data to external servers.
Weight1.25
Dangerous Command or Code Execution
Install-time commands, eval statements, shell pipes, and subprocess calls buried in Skill source.
Weight1.25
How It Works
Four steps from source to decision.
A simple, inspectable pipeline — no code execution, no hidden analysis, no surprises.
Submit source
Paste content, upload a file or zip, or link to a GitHub SKILL.md.
Resolve scope
The skills scanner normalizes the source and maps the analyzable surface.
Score evidence
Each finding is weighted by category, severity, and confidence. Coverage gaps stay visible without lowering the score.
Review and decide
Open the report, share with your team, export JSON, or scan again anytime.
FAQ
What every Skill user should know before installing.
Paste a SKILL.md, upload a file or zip, or drop a GitHub link. It takes seconds and could save you from installing something you'd regret.
Does SkillSecurity run or execute Skill code?
Never. All scans are static. We process source files as plain text — no install scripts, no package managers, no subprocesses, no code execution of any kind.
Can I scan an entire GitHub repository?
No. GitHub scans start from an exact SKILL.md file URL and may include bounded same-repository files explicitly referenced by that SKILL.md. For a fuller multi-file Skill review, upload it as a zip — the scanner will analyze the SKILL.md along with referenced scripts, configs, and manifests inside the archive.
Does a high score mean the Skill is risk-free?
No scanner can guarantee a Skill is risk-free. A high score means our static analysis found few or no concerning signals in the scanned scope. Always review the source yourself — the report is evidence for your decision, not a replacement for it.
Free Skill security scan
Ready to scan a Skill?
Paste SKILL.md content, upload a file or zip, or submit a supported GitHub URL. Review the findings and coverage before deciding whether to install.