ClawHub
Back to skill

Security audit

compliance-aiops

Security checks across malware telemetry and agentic risk

Overview

This skill is a local compliance-evidence tool that clearly discloses its audit-log reads, local bundle writes, and optional signing secret use.

Install only if you are comfortable letting the tool read local AIops audit databases and write compliance bundles/state under ~/.compliance-aiops. Use a dedicated account or filesystem permissions if you need query-only behavior, and treat generated bundles as evidence artifacts, not compliance certification.

SkillSpector

By NVIDIA
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 (2)

Missing User Warnings

Low
Confidence
76% confidence
Finding
The skill describes write-capable tools that create or modify local bundle files and signatures, but does not prominently warn users that these operations change the filesystem. In a compliance/evidence context, understated write behavior can mislead users into treating the skill as effectively read-only, increasing the risk of unintended artifact creation, overwrites, or accidental modification of evidentiary materials.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
## Authorization is not this tool's job — decide it where it belongs

Whether a bundle should be produced or signed is your decision, or the account's.
The tool does not gate it — there is no read-only switch and no approval prompt
to configure. The two right places to control it:

- **The account it runs as.** The tool only ever writes under
Confidence
93% confidence
Finding
The guardrail explicitly states that bundle production or signing is not gated by the tool and that there is no approval prompt or read-only mode, delegating authorization entirely to the runtime account or outer agent prompt. For a compliance evidence tool that can create signed, tamper-evident evidence bundles, this increases the risk of unauthorized or premature evidence generation, especially if the surrounding agent prompt is weak or the account is overprivileged.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.prompt_injection_instructions

Prompt-injection style instruction pattern detected.

Warn
Code
suspicious.prompt_injection_instructions
Location
references/agent-guardrails.md:45