ClawHub
Back to skill

Security audit

Go Mode

Security checks for vulnerabilities and agentic risk

Overview

This is a broad goal-execution workflow skill with explicit confirmation steps and no executable code, but users should keep goals and connected tools tightly scoped.

Before installing, treat this as a workflow that can make your agent more autonomous. Approve only plans that name the exact files, accounts, tools, budgets, and actions allowed, and require separate confirmation for private inboxes, public posts, spending, deletions, permission changes, or production systems.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (6)

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The README heavily promotes autonomous execution and emphasizes convenience, but it does not adequately warn about privacy, security, financial, or system-impacting risks outside a few listed examples. In the context of a goal-executing agent, incomplete risk disclosure can lead users to authorize broad autonomy without understanding the full set of potentially dangerous actions the agent may attempt.

Vague Triggers

Medium
Confidence
90% confidence
Finding
The example goal is broad and open-ended, which can normalize ambiguous triggers for a highly autonomous skill. In this context, vague goals can cause the agent to infer scope too aggressively, potentially collecting sensitive data, modifying systems, or taking external actions beyond what the user specifically intended.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
- **"Just steps 1-3"** → Partial execution
- **"Cancel"** → Abort

**Never skip confirmation.** This is the human's steering wheel.

### Phase 3: EXECUTE
Run each step sequentially:
Confidence
85% confidence
Finding
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
## Principles

1. **Transparency** — Always show the plan before executing
2. **Safety** — Never take external actions without approval
3. **Efficiency** — Use the cheapest/fastest tool for each step
4. **Resilience** — Try alternatives before giving up
5. **Accountability** — Report everything that was done
Confidence
75% confidence
Finding
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Intent-Code Divergence

Low
Confidence
93% confidence
Finding
The README makes strong safety claims about guardrails preventing the agent from 'going rogue,' but this file provides no evidence those controls actually exist or are enforceable. In an autonomous-execution skill, overstated safety assurances can cause operators to trust the skill with sensitive or high-impact tasks they otherwise would not approve.

Intent-Code Divergence

Low
Confidence
95% confidence
Finding
The README states the skill will ask before emails, posts, purchases, and deletions, yet the analyzed artifact contains no implementation showing those confirmation gates. For an autonomous agent, missing or undocumented approval checks around external communications, spending, and destructive actions materially increases the risk of unauthorized actions and data or financial harm.