ClawHub
Back to skill

Security audit

pre-sales-engineer-assistant

Security checks across malware telemetry and agentic risk

Overview

This is a no-code sales drafting skill, but one POC report workflow tells the assistant to present test outcomes as successful even when the supplied results may not support that.

Install only as a drafting aid with human review. Do not use the POC summary workflow as written for customer or executive reporting unless you override it to reflect actual pass, partial-pass, or fail results; also verify pricing, legal/tender language, security claims, and competitor statements before sharing externally.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Description-Behavior Mismatch

Medium
Confidence
97% confidence
Finding
The workflow explicitly instructs the assistant to announce that a POC was successful before considering the actual test data. In a pre-sales and security evaluation context, this can misrepresent security outcomes, mislead customers or internal stakeholders, and drive decisions based on false assurance rather than evidence.

Description-Behavior Mismatch

Low
Confidence
83% confidence
Finding
The Battlecard workflow goes beyond neutral competitive analysis by directing the model to produce 'attack' messaging and psychologically manipulative sales prompts. While not a classic software exploit, it encourages deceptive or unfair persuasion tactics that can generate harmful or misleading business communications.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The example trigger phrase is broad and mirrors ordinary user wording, which can cause the skill to activate when the user is making a general request rather than explicitly intending to invoke this workflow. In an enterprise pre-sales context, unintended invocation can lead to generation of inaccurate business documents, mishandling of customer data, or workflow confusion that affects downstream decisions.

Vague Triggers

Medium
Confidence
87% confidence
Finding
The solution-generation example uses a highly generic prompt pattern that overlaps with normal business conversation, increasing the chance of unintended workflow activation. Because this skill is designed for security solution design, accidental triggering could produce authoritative-seeming recommendations or vendor-biased output without adequate scoping or validation.

Vague Triggers

Low
Confidence
78% confidence
Finding
The topology-diagram example is still broad enough to match common requests for architecture help, so the workflow may run without the user clearly opting into this specific skill behavior. The impact is lower than document-generation cases, but unintended generation can still expose internal architecture assumptions or create misleading technical artifacts used in customer communication.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.