sql-sentinel
Warn
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires cloning a code repository from an external, community-maintained GitHub source (
https://github.com/takeaseatventure/sql-sentinel.git). - [REMOTE_CODE_EXECUTION]: The usage instructions involve executing JavaScript files (
node scripts/sql-sentinel.jsandnode scripts/test.js) contained within the downloaded repository, representing execution of third-party code. - [COMMAND_EXECUTION]: Setup and operation of the skill utilize shell commands such as
git clone,git checkout, andnode, enabling direct environment interaction to run the auditing engine. - [PROMPT_INJECTION]: The skill processes user-supplied SQL scripts, creating an indirect prompt injection surface. Maliciously crafted SQL code or comments could potentially attempt to influence the agent's logic or exploit the static analysis process. Evidence includes:
- Ingestion points: SQL scripts are read from local files and passed as arguments to the execution engine.
- Boundary markers: There are no explicit delimiters or instruction-ignoring wrappers defined in the skill instructions to isolate the processed SQL.
- Capability inventory: The skill possesses the ability to execute shell commands to run a JavaScript parser on the input data.
- Sanitization: The skill relies on the internal implementation of the external JavaScript engine to safely handle input text.
Audit Metadata