SQL Injection Testing
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEOBFUSCATIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [OBFUSCATION]: The skill documents multiple techniques to hide SQL keywords and parameters from security filters. This includes the use of hex-encoded strings (e.g., 0x61646D696E for 'admin') to evade string-based detection, the employment of URL encoding (%27, %22) and double URL encoding (%2527) for special characters, the use of inline SQL comments (/**/) and whitespace manipulation (%09, %0A) to break up recognizable patterns, and the implementation of case variation or double-writing (SELSELECTECT) to bypass keyword blacklists.\n- [DATA_EXFILTRATION]: Instructions include payloads for out-of-band (OOB) data extraction that send information to external domains not on the permitted whitelist. Examples include DNS exfiltration payloads targeting attacker.com and UNC path requests to attacker-server.com, as well as Oracle UTL_HTTP requests for data exfiltration.\n- [INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by directing the agent to interact with and process data from external web applications. \n
- Ingestion points: Target web application responses (SKILL.md).\n
- Boundary markers: Absent; no instructions are provided for delimiters or ignoring embedded instructions in external content.\n
- Capability inventory: The skill provides a comprehensive library of SQL injection payloads and guidance for tool usage (SKILL.md).\n
- Sanitization: Absent; no sanitization or escaping steps are defined for handling response data from target applications.\n- [COMMAND_EXECUTION]: The skill instructs the agent to use external command-line tools, such as SQLMap and Burp Suite, to automate the detection and exploitation of database vulnerabilities.
Audit Metadata