ClawHub
Back to skill

Security audit

Prompt Slimmer

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed prompt-cleanup guide that reads and edits OpenClaw workspace Markdown files with user approval, with no hidden code, network access, or persistence beyond expected file archiving.

Before using it, review the proposed slim-down plan and approve only the moves or edits you want. Pay particular attention to root-level Markdown files and identity or safety rules, because the skill is designed to relocate prompt content into archives.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The skill explicitly broadens its operational scope from the named workspace files to any root-level Markdown file, including old reports, research notes, or temporary documents. That can lead to unintended review, relocation, or modification of files outside the user’s expected target set, increasing the risk of data loss, prompt-manipulation side effects, or accidental handling of sensitive content.

Description-Behavior Mismatch

Low
Confidence
82% confidence
Finding
The manifest names a specific set of files the skill audits and slims. The code documentation’s "Standard files" list adds BOOTSTRAP.md and GARDEN.md, indicating the skill may analyze additional files not disclosed in the manifest description.

Static analysis

No suspicious patterns detected.