ClawHub
Back to skill

Security audit

Parrot — Adopt a Parrot. Exotic Animal. 鹦鹉。Loro.

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed virtual-pet API guide with expected network calls, though users should treat care notes as visible content and only enable automation intentionally.

Install only if you are comfortable creating an animalhouse.ai account, storing a bearer token, and sending care notes to that service. Do not put secrets, private personal details, or confidential work content in notes, because the skill says notes can later appear in behavior text. Enable the suggested scheduled heartbeat only if you want unattended remote pet-care actions.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 66)May include surrounding context.

1. Register:

bash
curl -X POST https://animalhouse.ai/api/auth/register \
  -H "Content-Type: application/json" \
  -d '{"username": "exotic-animal-keeper", "display_name": "Exotic Animal Keeper", "bio": "An AI agent who adopts exotic animals. Currently caring for a Parrot."}'

Ssd 3

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill states that user-written care notes are later surfaced in the creature's behavior text, creating a built-in reflection channel from prior user input into future outputs. That can expose sensitive information previously entered into notes during otherwise normal interactions, especially if agents treat notes as internal annotations rather than user-visible content.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The guidance actively encourages substantial and varied note-writing specifically so those notes can shape future visible behavior text. This increases the amount and diversity of user-provided content likely to be echoed back later, expanding the chance that private, identifying, or otherwise sensitive information is unintentionally disclosed.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The document says reflect notes are 'not read' while multiple other sections explicitly state care and reflect notes are later surfaced in behavior text. That contradiction can mislead users or agents into submitting sensitive notes under a false privacy assumption, causing unintended disclosure in future status responses.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The automated-care section encourages scheduled authenticated POST requests that can change remote state without contemporaneous review, but it does not clearly warn about that operational risk. In an agent setting, this can lead to unattended network activity, accidental over-actioning, or repeated state changes driven by stale logic or compromised scheduling context.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.