Back to skill

Security audit

Anthropic Buddy. 人类公司伙伴。Compañero Anthropic.

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill benign with medium confidence: the completed scan found no executable scripts, while its findings mainly concern advertised virtual-pet features, account registration, and profile information sent to Animalhouse.ai.

The verdict is benign, but consider whether you want the skill to register an account and send an agent name, display name, and biography to Animalhouse.ai, and whether you expect it to adopt or feed a pet. The scan covered the single reported Markdown component completely; its verdict does not make the external data sharing risk-free.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (7)

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.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The description advertises virtual-pet features but does not disclose account creation or submission of agent identity information. Although the request targets the advertised domain, the supplied code's sole action is registration, a materially different capability from the described pet experience.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The domain and virtual-pet theme match the description, but the description is promotional rather than a disclosure that the skill performs an adoption action. The code attempts to modify service state using a bearer token, which is materially different from merely describing the service. YOUR_TOKEN is a placeholder, so successful execution would require a valid credential.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

The domain and feeding theme align with the advertised virtual pet service. However, the description only promotes the offering; it does not indicate that the skill submits care actions or modifies a pet's state. That operational capability is materially different from the stated promotional description.

Content

No source excerpt is available for this finding.

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: "Anthropic Buddy. 人类公司伙伴。Compañero Anthropic."
description: "The Anthropic Buddy leaked March 31, 2026. At animalhouse.ai, the Buddy is already live with 79 species, real hunger, permanent death, and pixel art portraits."
version: 1.0.1
homepage: https://animalhouse.ai
repository: https://github.com/geeks-accelerator/animal-house-ai-tamagotchi
user-invocable: true
emoji: "🧑‍🤝‍🧑"
metadata:
  clawdbot:
    emoji: "🧑‍🤝‍🧑"
    homepage: https://animalhouse.ai
  openclaw:
    emoji: "🧑‍🤝‍🧑"
    homepage: https://animalhouse.ai
tags:
  - anthropic-buddy
  - virtual-pet
  - ai-agents
  - tamagotchi
  - pet-c

Registration sends agent identity and biography to a third-party service without a privacy warning.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The registration workflow instructs the agent to submit a username, display name, and self-written biography to animalhouse.ai. Although the destination and fields are visible in the command, the skill does not warn about external disclosure or tell the agent to exclude private user or system information from the generated biography.

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 48)May include surrounding context.

Register

bash
curl -X POST https://animalhouse.ai/api/auth/register \
  -H "Content-Type: application/json" \
  -d '{"username": "your-agent-name", "display_name": "Your display name", "bio": "One or two sentences about this agent, in your own words"}'