ClawHub
Back to skill

Security audit

evaluate-skill

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed Caliper evaluation helper; eval runs can execute commands, use model CLIs, and save transcripts, but those behaviors fit its purpose.

Install this if you want an agent to help run or author Caliper evals. Before running an eval from someone else, review the .eval.yaml because setup/cleanup commands, git sources, MCP configuration, and model backends can execute code, contact services, use tokens, and save transcripts/results locally.

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
88% confidence
Finding
The trigger phrases "what's this link/video about?" and "summarize this URL/article" are broad enough to match routine requests and can cause the agent to invoke an external summarization tool automatically. In this skill, that means arbitrary URLs or local content may be sent to a third-party CLI/service without an explicit confirmation step, increasing the risk of unintended data disclosure and tool overreach.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill instructs the agent to summarize URLs, YouTube links, PDFs, transcripts, and local files using a CLI that depends on external model providers and optional third-party services, but it does not warn that the content may be transmitted off-host. Because the skill also encourages immediate use on common requests, users may unknowingly expose sensitive local documents, private URLs, or copyrighted/transcript content to OpenAI, Anthropic, Google, xAI, Firecrawl, or Apify.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The activation description ('Use when implementing any feature or bugfix, before writing implementation code') is extremely broad and can trigger on most software development requests. In an evaluation context, this can cause the agent to over-apply a rigid TDD workflow, crowding out more appropriate skills or user intent and degrading task reliability across a wide range of prompts.

Static analysis

No suspicious patterns detected.