filtering-query-builder

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONNO_CODE
Full Analysis
  • [NO_CODE] (INFO): The skill consists solely of a markdown manifest (SKILL.md) and contains no executable scripts or source code for analysis.
  • [COMMAND_EXECUTION] (HIGH): The skill requests permission to use Bash(curl:*). This provides a broad capability to execute shell commands and perform network operations without restriction on the destination or the commands run by the agent.
  • [INDIRECT_PROMPT_INJECTION] (HIGH): The skill is designed to be triggered by external user requests regarding API development. Mandatory Evidence: 1. Ingestion points: User requests mentioning 'filtering query builder' or asking about API patterns (documented in SKILL.md). 2. Boundary markers: None defined to isolate user input from tool arguments. 3. Capability inventory: Permission to use Write, Edit, and Bash(curl:*) across all interactions. 4. Sanitization: No sanitization logic is present to prevent malicious instructions in user queries from being passed directly to the shell or file-writing tools. The combination of high-privilege tool access and untrusted input processing creates a high-severity vulnerability surface.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 12:50 PM