ClawHub
Back to skill

Security audit

Reframe

Security checks across malware telemetry and agentic risk

Overview

This is a simple reframing skill, but it also tells the agent to invoke another skill to repair or optimize itself without clear user authorization.

Review before installing. The summarization workflow itself is low risk, but the self-optimization and skill-manager instructions should be removed or constrained to an explicit, user-approved maintenance task.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The skill instructs the agent to load a separate 'skill-manager' to repair or update this skill when verification fails. That creates unjustified cross-skill delegation and opens a path for unbounded behavior changes outside the declared purpose of simple problem reframing, increasing the attack surface and enabling prompt-chain escalation.

Context-Inappropriate Capability

Medium
Confidence
97% confidence
Finding
The 'self-optimization' and 'self-evolution' language authorizes the skill to modify itself or seek optimization even when no user asked for that behavior. Self-referential improvement instructions are dangerous because they can override scope boundaries, trigger recursive tool/skill use, and normalize unauthorized prompt mutation.

Vague Triggers

Medium
Confidence
90% confidence
Finding
The trigger conditions are broad and mandatory, covering subjective states like 'too complex,' 'no progress,' or any request to rethink. Overly broad auto-invocation can cause the skill to activate in unrelated contexts, unexpectedly write/read temp files, and steer the agent away from the user's intended workflow.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The trigger list contains common conversational phrases such as 'simple,' 'another way,' or 'forget it,' which are likely to appear in ordinary dialogue without intending skill activation. This increases accidental invocation risk and can produce unnecessary context compression or workflow diversion.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.