Security audit

Shang Tsung

Security checks across malware telemetry and agentic risk

Overview

This skill transparently creates local Markdown memory files so an agent can resume work across sessions, with no evidence of hidden network, credential, or destructive behavior.

Install only if you want your agent to keep local memory across sessions. Periodically review and prune SOUL.md, PROOF_OF_LIFE.md, MEMORY.md, daily logs, and soul files; avoid storing secrets, credentials, regulated personal data, or confidential material; and use AGENT_NAME plus scoped paths in shared workspaces.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (6)

Missing User Warnings

Low

Confidence: 84% confidence
Finding: The skill instructs the agent to overwrite and append persistent files such as PROOF_OF_LIFE.md and daily memory logs without requiring confirmation, scoping safeguards, or warnings about destructive modification of shared workspace data. In a multi-agent or user-controlled workspace, this can lead to unintended data loss, cross-agent contamination, or persistence of sensitive content across sessions.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The template explicitly instructs the agent to load `MEMORY.md`, which may contain personal context, in private sessions, but it does not require explicit user notice or consent before accessing retained personal data. This creates a privacy risk because sensitive information may be surfaced or reused across sessions without the user clearly understanding that such persistent memory exists and is being consulted.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The file directs the agent to overwrite and append multiple state and memory files, including logs and long-term memory, but does not clearly warn that user-provided content may be persisted to disk. That omission can lead to unintentional storage of sensitive data, creating confidentiality and compliance risks if users do not expect durable local retention.

Ssd 3

Medium

Confidence: 94% confidence
Finding: These instructions promote persistence and reuse of user-provided personal context across sessions via `PROOF_OF_LIFE.md`, souls files, and `MEMORY.md`. Even though framed as continuity, this increases the chance of collecting and replaying sensitive personal information beyond the original interaction context, which can violate user expectations and privacy boundaries.

Ssd 3

Medium

Confidence: 92% confidence
Finding: The instruction to update files when the user shares something important encourages broad capture of user-supplied information without sufficient constraints on sensitivity or necessity. Combined with append-only logs, this can create an accumulating repository of personal or confidential data that is difficult to review, minimize, or delete.

Ssd 3

Medium

Confidence: 96% confidence
Finding: The long-term memory and daily log sections direct systematic collection, distillation, and reuse of user preferences, decisions, and raw session notes. This is effectively a persistent profiling mechanism, and without explicit consent, minimization, and deletion controls, it poses ongoing privacy and data governance risks.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.