The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The instructions guide the agent to download and run third-party code. Specifically, it suggests using git clone followed by node scripts/sql-sentinel.js to execute the analysis logic.
[EXTERNAL_DOWNLOADS]: The skill relies on fetching the core executable engine from an external repository (https://github.com/takeaseatventure/sql-sentinel.git) that is not part of the recognized trusted sources. While the documentation includes a warning to review specific commits, the default path involves downloading unvetted code.
[COMMAND_EXECUTION]: The skill provides shell command patterns for the agent to execute, including repository cloning and script execution via Node.js.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided SQL scripts. This creates a surface for indirect prompt injection if those scripts contain malicious instructions hidden in comments or metadata designed to influence the agent's interpretation of the audit results.
Ingestion points: Processes .sql files and SQL strings provided by users.
Boundary markers: None explicitly mentioned to separate untrusted SQL content from the agent's instructions.
Capability inventory: Executes shell commands, clones remote repositories, and runs local JavaScript processes via node.
Sanitization: No evidence of sanitization or filtering for the SQL content being analyzed.

sql-sentinel