ClawHub
Back to skill

Security audit

pipeline-architecture

Security checks across malware telemetry and agentic risk

Overview

This is a non-executable architecture guide, but it uses very broad mandatory activation rules that can steer many unrelated coding tasks without clear user opt-in.

Install this only for projects where you explicitly want all business-logic changes to follow this pipeline architecture. Be cautious in general-purpose workspaces because the broad trigger terms may cause the agent to apply this pattern during normal refactors or tests. For real file, device, or external API mutations, add project-specific allowlists, authorization checks, idempotency safeguards, and rollback or compensation rules before using the generated pattern in production.

SkillSpector

By NVIDIA
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 (4)

Context-Inappropriate Capability

Medium
Confidence
72% confidence
Finding
The reference explicitly includes a mutation path for device or host variables in a broadly reusable architecture skill without defining safety boundaries, authorization requirements, or target restrictions. In an agent-assisted coding context, this can normalize generation of code that modifies system/device state, increasing the chance of unsafe persistence adapters being implemented and later abused for configuration tampering or host-impacting changes.

Vague Triggers

High
Confidence
97% confidence
Finding
The skill declares extremely broad activation keywords such as service, pipeline, step, query, mutation, refactor, test, and even generic code-smell terms. These collide with ordinary development requests, making the agent likely to load and obey this skill in many unrelated contexts, which can silently override user intent and steer code generation toward a mandated architecture.

Natural-Language Policy Violations

Medium
Confidence
90% confidence
Finding
The skill instructs the agent that any business-logic code must follow this architecture and that it must load language-specific references based on detected project language, rather than waiting for user direction. This creates a coercive behavior where the skill can dictate implementation structure and language handling without explicit opt-in, reducing user control and increasing the chance of inappropriate or policy-conflicting guidance.

Missing User Warnings

Medium
Confidence
84% confidence
Finding
The document notes that external API mutations are not rolled back, but it understates the consequence that data may already have been transmitted to third parties before a later pipeline failure. That omission can lead developers to overestimate atomicity and privacy guarantees, creating real risks of data leakage, duplicate side effects, or inconsistent cross-system state.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.