ClawHub
Back to skill

Security audit

tldr

Security checks for vulnerabilities and agentic risk

Overview

This skill only changes response formatting to put a TL;DR first and does not add execution, data access, persistence, or hidden behavior.

Install this if you want answers to start with concise summaries. Be aware it can keep changing response style for the rest of a session until turned off, and broad phrases like 'more' or 'normal mode' may be interpreted as formatting controls.

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 (4)

Vague Triggers

Medium
Confidence
90% confidence
Finding
The manifest describes the command as turning on TL;DR output or compressing 'one target,' and the argument is described as 'a thing to compress.' This is ambiguous and lacks clear trigger constraints or exclusion conditions, which could cause unintended invocation from generic summarization-style requests.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The skill defines broad deactivation phrases like "normal mode" and activation persistence for the rest of the session without strong scoping. An unrelated user message containing one of these phrases could silently disable or alter the skill's behavior, causing loss of expected formatting and potentially suppressing always-visible safety-relevant content such as security findings or destructive-action warnings.

Vague Triggers

Medium
Confidence
94% confidence
Finding
The brief-mode trigger treats broad follow-up language like "more," "expand," and even a generic follow-up question about the same answer as a control signal. This can be activated unintentionally during ordinary conversation, causing the assistant to reveal deferred content or change response shape without a clear user intent boundary; in multi-agent or tool-mediated settings, such ambiguity can also be abused to manipulate output flow.

Rp1

Medium
Category
MCP Rug Pull
Confidence
75% confidence
Finding
Docker image references without a specific tag (:latest is implicit) or digest (@sha256:...) can be silently replaced by a malicious image.

Static analysis

No suspicious patterns detected.