ClawHub
Back to skill

Security audit

Advanced Evaluation

Security checks for vulnerabilities and agentic risk

Overview

This is a documentation-only skill for LLM output evaluation, with minor caution around broad activation wording and requests for detailed evaluator rationales.

This skill appears safe to install as guidance material. Before using its prompt templates, narrow activation to LLM evaluation tasks and adapt the scoring prompts to ask for brief evidence-backed rationales rather than detailed chain-of-thought.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

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 (2)

Vague Triggers

Medium
Confidence
88% confidence
Finding
The manifest description says the skill should activate when the user mentions broad phrases like "compare model outputs," "create evaluation rubrics," or general concepts such as "automated quality assessment." These triggers are not tightly scoped and no exclusion conditions or negative examples are provided, which could cause the skill to activate in contexts beyond the intended advanced LLM-as-judge use case.

Context-Inappropriate Capability

Medium
Confidence
84% confidence
Finding
The skill explicitly requires chain-of-thought-style justification before scoring, which asks downstream evaluators to elicit detailed internal reasoning rather than concise, evidence-based outputs. In an evaluation skill, this is unnecessary for core functionality and can increase prompt leakage, overcollection of sensitive reasoning traces, and non-compliance with safer best practices that prefer brief rationales or evidence summaries.

Static analysis

No suspicious patterns detected.