ClawHub
Back to skill

Security audit

Cognition

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent local memory system, but its persistent future-intent template can auto-execute broad deferred actions without a clear approval gate.

Install only if you want the agent to maintain durable local memory. Before enabling FUTURE_INTENTS behavior, revise it so deferred actions require explicit approval, narrow triggers, expiration, and auditability; otherwise actions may run later based on vague context.

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
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The template defines triggers such as time-based, event-based, and especially context-based activation in broad natural language terms, then instructs that intents are scanned every session start and every heartbeat and executed when triggered. That combination can cause deferred actions to fire unexpectedly or too often, creating a persistent instruction channel that may override current user intent or cause unauthorized actions based on vague conditions like 'when topic arises'.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The prompt explicitly instructs the workflow to write to memory/summaries/YYYY-WNN.md, which is a durable memory location, but it does not provide a clear user-facing warning or approval gate for that persistence. Even though the content is framed as analysis-only and avoids editing higher-risk files, it still causes lasting state changes that could surprise users or create unreviewed memory accumulation.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.