ClawHub
Back to skill

Security audit

OpenClaw Knowledge Base Retriever

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local knowledge-base search helper with disclosed, bounded file access and optional user-confirmed maintenance writes inside the chosen knowledge base.

Install only if you want an agent to read files under a knowledge-base directory you name. Review prompts carefully before allowing index generation, PDF text/image extraction, or Python package installation, and prefer the packaged scripts over copying raw examples from the reference docs.

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 (3)

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The document first establishes that PDF extraction and any file writes must go through packaged helper scripts with path-boundary enforcement, then later provides direct code examples that read arbitrary PDF paths and write output files without those safeguards. In an agent-skill context, example code is often copied verbatim, so these snippets can bypass the intended symlink/root containment controls and lead to unauthorized reads or writes if an agent or maintainer follows them instead of the guarded scripts.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The manifest repeatedly emphasizes a local, non-networked knowledge-base retriever and says the skill itself does not network or send data. However, the SKILL file prominently directs the user/agent to a GitHub repository and vendor website, which is behavior outside the core local-retrieval scope described in the manifest.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The platform strategy and decision table present pypdfium2 as the standardized rendering path for scanned PDFs, intended to avoid platform differences. However, the OCR example at L167-L180 uses pdf2image.convert_from_path, which is a different implementation path and undercuts the stated 'unified Python route' guidance.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.