ClawHub
Back to skill

Security audit

Triage / Task Prioritization

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed task-prioritization guide that stores only confirmed local preferences and does not contain code, network retrieval, credential handling, or hidden execution.

Before installing, expect this skill to influence task ordering and interruption decisions, and only allow it to save standing priority rules when you explicitly agree. Be careful with broad urgency phrases if you do not want triage applied to a casual request.

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
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Vague Triggers

Medium
Confidence
92% confidence
Finding
The skill advertises activation on very common phrases like 'urgent', 'ASAP', 'drop everything', and 'no rush', which are likely to appear in ordinary conversation outside an explicit triage request. That can cause unintended routing into this skill, leading the agent to reprioritize work or interrupt current context when the user was only speaking casually. Because this skill can persist user preference data and influence task ordering, accidental activation has more than cosmetic impact.

Vague Triggers

Medium
Confidence
90% confidence
Finding
The trigger guidance says the skill should engage when the user uses priority language, but it does not sufficiently constrain whether that language refers to actual task-prioritization intent. Words like 'urgent' or 'no rush' are highly ambiguous and may describe only one item, tone, or narrative context, causing the model to invoke triage logic when it should simply answer the underlying request. In this skill, that increases the chance of unnecessary queue reshuffling, interruptions, and mistaken learning from incidental phrasing.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
Ask what consumes the output. The meeting, decision, launch, or dependent person behind the deadline is the real one; the stated date is often that minus someone's buffer.

Soft-deadline tells (each lowers confidence in the date; none abolishes it without checking):

- Round dates with no named consumer ("by the 1st", "end of month").
- Habitual "EOD" from a sender whose EODs never had consequences — sender calibration applies (signals.md).
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.

Static analysis

No suspicious patterns detected.