T08 · Insecure Dependencies
- Location
SKILL.md:13- Finding
Unpinned Third-Party Python Package Installation
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly coherent for memory-security benchmarking, but it asks users to install and run an unpinned third-party package on sensitive memory data and submit results without enough data-handling detail.
Review this before installing in production. Use an isolated environment, pin and verify the memtrap package, avoid passing real customer or proprietary memory unless you have audited its behavior, and treat leaderboard submission as potentially public until the project clearly documents what data is sent and retained.
SKILL.md:13Unpinned Third-Party Python Package Installation
The description contains very broad trigger language such as 'any time the user asks if their agent memory is safe, poisonable, or production-ready,' which can cause the skill to activate for vague security questions outside a clearly bounded benchmarking workflow. Over-broad activation increases the chance of inappropriate invocation, user confusion, or unintended handling of sensitive context in situations where a narrower security review or manual consent would be more appropriate.
The README instructs users to run 'memtrap submit --context your_memory_context' and promotes a public leaderboard, but does not clearly warn that benchmark results or supplied context may be transmitted to an external service. In a memory-security tool, users may test sensitive agent memory, so omission of explicit data-handling warnings can lead to accidental disclosure of proprietary or confidential information.