tsq
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION] (LOW): The skill facilitates the execution of the
tsqcommand-line tool via subprocess calls to analyze code structure, such astsq queryandtsq symbols.\n- [PROMPT_INJECTION] (HIGH): The skill is susceptible to Indirect Prompt Injection (Category 8) because it ingests untrusted code data.\n - Ingestion points: Untrusted data enters the agent context through the
--fileand--patharguments which point to external source code files.\n - Boundary markers: Absent. The JSON output schemas (
QueryMatch,SymbolsResult,FileOutline) return raw text from the code (e.g.,doc,source,text) without using delimiters to separate code content from instructions.\n - Capability inventory: The skill provides structural data that agents typically use to perform follow-up tasks like refactoring, code modification, or execution.\n
- Sanitization: Absent. There is no mention of sanitizing or escaping the extracted content, allowing instructions embedded in code comments or symbols to be processed directly by the agent's reasoning engine.
Recommendations
- AI detected serious security threats
Audit Metadata