debugging-orm-queries
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection] (HIGH): The skill processes untrusted database logs and EXPLAIN output, which is a significant injection surface. Ingestion points:
queries.logand database EXPLAIN plans as described inSKILL.mdandevals/evaluations.json. Boundary markers: Absent. Capability inventory: File writing (fs.writeFileinlog_interceptor.js) and shell command execution. Sanitization: Absent. - [Command Execution] (MEDIUM): Documentation in
SKILL.mdencourages piping database CLI output (e.g.,psql,mysql) directly into Python scripts, which could be exploited if the output contains malicious content. - [File Access] (MEDIUM): The
log_interceptor.jsscript includes anexportLogsfunction that writes query data to an unvalidatedfilepathprovided as an argument, allowing for potential arbitrary file creation or overwriting. - [Unverifiable Dependencies] (MEDIUM):
SKILL.mdreferencesquery_analyzer.pyandexplain_parser.py, but the source code for these scripts is not included in the provided package.
Audit Metadata