TurboQuant Optimizer

• Verdict: Pass
• Confidence: Medium
• Analyzed: Apr 15, 2026, 2:40 PM
• Findings: 0
• Version: 1.0.0

Overview

The skill's code, docs, and runtime instructions are consistent with a token-optimization tool that intercepts and compresses OpenClaw conversation data, but it reads and caches user session data so you should audit where data is stored and run it in a sandbox first.

This skill appears to be what it claims: a conversation optimizer that reads sessions, compresses context, and caches results. However, before installing or enabling it on production data: 1) Audit where checkpoints/caches are stored (search for cache paths) and confirm retention, encryption, and deletion behavior. 2) Run it in a sandbox or dev agent to observe filesystem and network activity (ensure it does not exfiltrate data). 3) Review the omitted/truncated library files if possible (the package includes several large source files; ensure no hidden networking or credential-reading code). 4) Because it hooks into pre-api-call, consider enabling it only for specific agents or disabling automatic invocation until you’re comfortable. 5) Verify the repository and maintainer contact (MincoSoft links are provided) and run the included tests/benchmarks (npm test, benchmark) locally. These steps will reduce risk from accidental exposure of chat contents or unexpected persistent storage.