ClawHub
Back to skill

Security audit

Drawio Diagram Builder

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed draw.io diagram-building helper with local validation and preview scripts; its network and file-write behavior fits that purpose.

Before installing, be aware that previews load diagrams through diagrams.net and may expose diagram contents to that service. For large or high-fidelity requests, set a clear iteration or time budget because the skill strongly encourages repeated refinement until quality gates are met.

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)

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
**Handoff thresholds (canonical — all files use these exact values):**

- **BLOCKED** — TOTAL < 30, OR any dimension ≤ 4. You must continue iterating. Do not ask the user — just keep working.
- **BORDERLINE** — 30 ≤ TOTAL < 40 with no dimension ≤ 4. List 5+ specific things you would fix in the next pass. Handoff only if the user explicitly asked for a quick result.
- **ALLOWED** — TOTAL ≥ 40 AND every dimension ≥ 6. A dimension scored 5 is NOT allowed — it is borderline; review and improve it before handoff.
Confidence
88% confidence
Finding
The instruction 'Do not ask the user — just keep working' encourages the agent to continue iterating autonomously rather than seeking clarification or confirmation when quality thresholds are not met. In an agentic environment, this can lead to unbounded tool use, excessive compute consumption, or acting beyond the user's intended stop condition, especially if combined with rendering or file-editing loops.

Static analysis

No suspicious patterns detected.