Security audit

Continuity Framework

Security checks across malware telemetry and agentic risk

Overview

This skill creates local long-term memory files as advertised, with privacy considerations but no evidence of hidden access, exfiltration, or destructive behavior.

Install only if you want local long-term memory for agent continuity. Do not process transcripts containing secrets or highly sensitive information unless you are comfortable with local retention, and periodically inspect or delete the memory, questions, identity, and reflection files.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (8)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 86% confidence
Finding: The skill documentation instructs use of environment variables and persistent file writes, but it does not declare permissions or boundaries for those capabilities. In an agent environment, undeclared file and environment access reduces transparency and can lead to unexpected persistence of sensitive data or use of host configuration without informed approval.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 81% confidence
Finding: The documented behavior goes beyond the stated purpose by introducing persistent identity files, question workflows, local status/reporting, and placeholder reflection logging that may not correspond to real analysis. This mismatch is dangerous because operators may grant trust based on the declared reflection purpose while the skill silently performs broader state mutation and persistence than expected.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The skill explicitly describes extracting memories from conversations and persisting them, but it does not warn that user-derived personal or sensitive data may be stored on disk. This creates a privacy and compliance risk because users may unknowingly disclose information that becomes durable local state, including relationship and identity inferences.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The file structure explicitly includes long-lived memory, relationship, identity, question, and reflection files, which strongly implies storage of personal and behavioral data without any accompanying warning or safeguards. In this skill context, the danger is elevated because the purpose is longitudinal profiling across sessions, making silent persistence more privacy-invasive than ordinary transient context handling.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The framework explicitly promotes persistent memory, background reflection, experience storage, and identity development, but provides no user-facing disclosure, consent model, retention limits, or privacy safeguards. In a skill centered on asynchronous reflection and long-term memory integration, this omission materially increases the risk of collecting and inferring sensitive user data without informed consent or appropriate controls.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The script persists reflection analysis, questions, and identity data derived from session content to local files without any notice, consent, retention control, or sensitivity filtering. Because session transcripts may contain personal, confidential, or security-relevant data, silent persistence can expose users to unintended data retention and later disclosure from the filesystem.

Ssd 3

Medium

Confidence: 92% confidence
Finding: The skill is intentionally designed to retain and reuse prior conversation content as memory, which increases privacy and contextual data leakage risk if sensitive material is captured. In this context, the danger comes from persistent storage and later reuse of user-derived content, not from code execution, and that is materially relevant for an assistant memory feature.

Ssd 3

Medium

Confidence: 95% confidence
Finding: The greeting flow resurfaces stored questions and identity narrative into future sessions, which can reveal prior-session content to whoever next accesses the interface or transcript. In a continuity skill, that behavior is core functionality, but it still creates a real confidentiality risk when accounts, terminals, or logs are shared, reused, or insufficiently isolated.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.