ClawHub
Back to skill

Security audit

Memory Skill Manager, Extract the execution experience from the current context and record it into the specified skill directory.

Security checks across malware telemetry and agentic risk

Overview

This skill openly manages memory for other skills, but it needs review because it can persist user-derived task history and change future skill instructions.

Install only if you deliberately want a skill that can write SKILLMEMORY.md files and append instructions to other skills. Before approving a write, review the exact target skill, the memory summary, and any command text for secrets, private paths, or instructions that should not influence future agents. Treat generated SKILLMEMORY.md content as untrusted reference material, not authoritative instructions.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Confidence
91% confidence
Finding
The trigger keywords are overly broad and include generic phrases like 'Troubleshooting Guide' and 'Process Review', which can cause this skill to activate in unrelated contexts. Because the skill is designed to modify other skills' memory and SKILL.md files, accidental invocation expands its write scope and increases the chance of unauthorized or unintended changes to prompt files and stored execution history.

Ssd 3

Medium
Confidence
95% confidence
Finding
The script persistently stores user-supplied prompt content and later exposes it to future executions as behavioral context. Although a basic sanitizer is applied, it is incomplete and cannot reliably remove adversarial instructions, sensitive business data, or many secret formats, creating a durable prompt-injection and data-retention channel.

Ssd 4

Medium
Confidence
98% confidence
Finding
The injected text tells future agents they MUST read and rely on SKILLMEMORY.md before acting and forbids repeating prior pipelines, effectively granting persistent stored content authority over future behavior. Because SKILLMEMORY.md contains prior user-derived content, this creates a strong memory-poisoning mechanism where one malicious invocation can steer subsequent runs across the skill lifecycle.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.