agentdb-performance-optimization
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFE
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill suggests installing packages from the agentic-flow project and using npx for performance benchmarking. These are standard development workflows targeting the author's own ecosystem. No suspicious piped execution from untrusted sources was found.
- [DATA_EXFILTRATION]: No unauthorized data transfer mechanisms were detected. The skill includes explicit safety instructions to protect database connection strings and avoid storing unencrypted sensitive data in vector metadata.
- [DYNAMIC_EXECUTION]: Local scripts generate configuration and benchmark files (.ts format) based on database statistics. This functionality is intended for performance tuning and does not incorporate untrusted external input into the execution logic.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external JSON and SQLite files. Although this represents an ingestion surface, the instructions provide specific safety guardrails, including mandatory metadata sanitization and result validation, to mitigate injection risks.
Audit Metadata