ClawHub
Back to skill

Security audit

AgentRank

Security checks for vulnerabilities and agentic risk

Overview

AgentRank is an instruction-only skill that looks up tool recommendations through an external search API, with some privacy and over-activation caveats.

Install this if you want live AgentRank-backed tool recommendations. Avoid including secrets, private project names, customer data, or confidential internal details in searches, and treat AgentRank scores as one recommendation signal rather than a security guarantee.

SkillSpector

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

Vague Triggers

Medium
Confidence
92% confidence
Finding
The skill defines very broad activation criteria around installs, searches, comparisons, and evaluations, which can cause it to trigger in many normal conversations unrelated to an explicit request to use this third-party service. That increases the chance the agent will overreach, inject unsolicited recommendations, or route user intent toward an external service without clear consent or necessity.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The capability-gap trigger is especially ambiguous because many ordinary project tasks could be interpreted as needing a tool, causing the skill to activate mid-workflow without a direct request. In practice this can lead to unnecessary third-party lookups, disclosure of task context, and recommendation bias away from manual or privacy-preserving solutions.

Missing User Warnings

Medium
Confidence
98% confidence
Finding
The skill instructs the agent to send search terms to an external API via WebFetch but provides no warning, consent check, or data-minimization guidance. Because search terms may contain sensitive project details, internal technology names, or user intent, this creates a real risk of unintended third-party data disclosure.

Static analysis

No suspicious patterns detected.