ClawHub
Back to skill

Security audit

Customer

Security checks for vulnerabilities and agentic risk

Overview

This is a plain-text customer success coaching skill with no code execution or persistence, but users should avoid pasting sensitive customer data unless their environment is approved for it.

Install is reasonable for customer success work. Before using it with real accounts, follow your company's data handling rules and prefer anonymized account labels, aggregated metrics, and redacted comments unless you are using an approved secure AI environment.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
User: I have a customer who has been using our analytics product for 10 months. Their NPS is 9. They have 15 active users out of 20 licensed seats. I want to start an expansion conversation but I do not know how to bring it up without seeming like I am just trying to sell them more.

Agent: The timing and the approach are everything here. Do not lead with the product you want to sell. Lead with curiosity about their situation. In your next check-in, after reviewing their current results, ask: "You have mentioned [specific outcome they achieved]. Is that outcome being felt across the whole team or mostly by the core group using the product?" That question surfaces whether other teams have the same problem without asking them to buy anything. If they say other teams are dealing with the same challenge, you have an organic opening: "That is actually something we see often — would it be useful to do a quick discovery call with [other team lead] to see if there is a fit?" You are not pitching. You are offering to solve a problem they just confirmed exists. The expansion conversation that starts from their problem closes at a much higher rate than the one that starts from your quota.

---
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.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The prompt encourages users to paste detailed customer portfolio data including health scores, renewal dates, and last-contact information without any guidance on redaction, minimization, or approved data handling. In a customer success context, this can expose confidential customer and business-sensitive information to the agent unnecessarily, increasing privacy, contractual, and data-governance risk.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
This section asks for customer-specific usage data and onboarding outcomes to prepare business reviews, which may include proprietary usage patterns, success criteria, and potentially personal or commercially sensitive information. Because the skill gives no warning or boundaries, it normalizes oversharing of sensitive customer data into the AI workflow.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
Requesting a full customer portfolio for quarterly analysis invites bulk disclosure of churn, expansion, NPS, and health distribution data, which can be highly confidential at the company and customer level. The danger is amplified by the broad scope of data requested and the absence of confidentiality, least-privilege, or minimization guidance.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The trigger list contains broad, common terms such as "customer," "renewal," and "onboarding" that are likely to appear in many unrelated conversations. This can cause unintended invocation of the skill, leading to context switching, noisy behavior, and possible inappropriate access to customer-oriented workflows when the user did not intend to use this skill.

Static analysis

No suspicious patterns detected.