Sample work · AI & Automation
Skill Security Audit: Detecting malicious patterns before installation
A demo skill in both a suspicious and a clean version. The auditor reads all files and detects injection, download-and-execute, exfiltration of environment variables, access to key files, code from external text, invisible characters, and excessive permissions. Nothing is installed or executed.
Sample work (demo), not a client job. Built as a trial task with invented example data; the check figures below come from the automatic test run and are not polished.
What was built
- All files read, not just the description
- Findings by severity with file, install verdict or not
- The test patterns are not executable, nothing is run
- Skill and operation are fictional
How it was checked
Automatic test run on 2026-10-31, taken from the log (test data invented, figures measured):
- Demo skill: suspicious version 10 findings in 8 groups (injection, download-and-execute, exfiltration, key files, code from external text, invisible characters, excessive permissions), clean version 0
- nothing installed or executed
- 11 checks
Limits
What this sample does not cover, so nobody expects more:
- Finds known patterns, no new tricks; obfuscated code only partially
- A clean finding is no guarantee
- Additionally check skills with write or network permissions in an isolated environment
Something like this for your business?
Describe briefly what you need. You get a fixed price before I start; nothing happens without your yes.