ClawHub
Back to skill

Security audit

AI Knowledge Workbench Lite

Security checks for vulnerabilities and agentic risk

Overview

This skill coherently builds a local, metadata-only knowledge index and dashboard in a user-authorized workspace, with no evidence of external upload, hidden persistence, or destructive behavior.

Install only for a workspace or vault you are comfortable indexing locally. The skill is designed not to upload note bodies or alter source notes, but it will create local metadata indexes and dashboard files, including relative paths, tags, headings, links, hashes, and diagnostic environment details in command output.

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (1)

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
The diagnosis result returns host fingerprinting details such as OS release, machine architecture, shell path, and whether the workspace path contains non-ASCII characters, even though the skill is described as a metadata-only local knowledge indexer. Collecting and surfacing extra host identity data increases unnecessary exposure to the calling agent or logs, and broadens the privacy and reconnaissance value of the output beyond what is needed for core functionality.

Static analysis

No suspicious patterns detected.