SQL Injection Testing

Fail

Audited by Gen Agent Trust Hub on Nov 6, 2026

Risk Level: HIGHPROMPT_INJECTIONDATA_EXFILTRATIONOBFUSCATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill's core purpose is to provide instructions and payloads for bypassing authentication mechanisms and exploiting web application vulnerabilities. The instructions explicitly direct the agent to perform actions like 'bypass authentication' and 'exploit database query vulnerabilities', which contradict standard safety alignment for AI agents.
  • [DATA_EXFILTRATION]: The documentation includes functional SQL payloads designed for Out-of-Band (OOB) data exfiltration to external, non-whitelisted domains such as 'attacker.com' and 'attacker-server.com'. These payloads are intended to leak database information via DNS or HTTP requests to infrastructure controlled by an adversary.
  • [OBFUSCATION]: The skill teaches and provides examples of obfuscation techniques used to evade security filters (WAF/IPS), including Hexadecimal encoding (e.g., '0x61646D696E'), URL encoding, and Double URL encoding. It also details the use of inline comments and whitespace manipulation to hide SQL keywords from detection systems.
  • [COMMAND_EXECUTION]: The workflow requires the installation and use of external automated exploitation tools, specifically 'SQLMap', which involves the execution of arbitrary commands and automated vulnerability probing from the host system.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a high-risk attack surface by interacting with untrusted external content.
  • Ingestion points: The skill ingests untrusted data from target web application URLs, URL parameters, form fields, and HTTP headers like 'User-Agent' and 'X-Forwarded-For'.
  • Boundary markers: There are no boundary markers or instructions provided to the agent to ignore or sanitize potentially malicious instructions embedded within the data retrieved from the target applications.
  • Capability inventory: The skill facilitates the execution of complex network operations (OOB exfiltration) and database interactions that can be triggered by external data.
  • Sanitization: The documentation lacks instructions for sanitizing user-provided URLs or validating the integrity of data received from the target before processing it.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Nov 6, 2026, 03:30 PM