sql-injection-prevention
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it is designed to ingest and process untrusted user-supplied PHP code for the purpose of refactoring (SKILL.md).
- Ingestion points: The skill's primary workflow involves reviewing and fixing user-provided code snippets containing SQL queries.
- Boundary markers: The instructions do not explicitly mandate the use of boundary markers or delimiters when interpolating untrusted code into the agent's context.
- Capability inventory: The skill enables the agent to perform code refactoring and generate SQL-safe code outputs.
- Sanitization: No specific input sanitization or verification logic is defined for the code being analyzed.
- [SAFE]: The core logic of the skill promotes defensive security practices such as using prepared statements with bound parameters, disabling emulated prepares in PDO, and implementing allow-lists for dynamic identifiers. No evidence of data exfiltration, obfuscation, or remote code execution was found.
Audit Metadata