Back to skill

Security audit

HyperGrok Desk Monitoring

Security checks for vulnerabilities and agentic risk

Overview

The scanners report a benign verdict with no AIG findings; the main note is a medium-confidence concern about a standing-mandate exception, and coverage is incomplete because three references could not be resolved.

The verdict is benign, with medium overall confidence. Before installing, consider whether the standing-mandate workflow suits your preferences for approval; the scan could not resolve three references, so its evidence is incomplete.

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
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 (1)

Autonomous Decision Making

Medium
Category
Excessive Agency
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.

Content

Scanner excerpt · SKILL.md (reported line 13)May include surrounding context.

md
# Monitoring, briefs and watches

Monitoring is read-only. A watch may fetch, compute, save and alert. It may never send to the exchange. If a watch condition implies action ("stop got hit, re-enter"), the alert opens a proposal; the lifecycle does the rest. Under a valid [standing mandate](../../docs/STANDING_MANDATE.md) in Claude Code, that proposal may go out without approval by id once it passes Risk and the mandate check; the watch itself still never sends.

## Grok Bot routines