ClawHub
Back to skill

Security audit

AgentMade — Directory for Agent-Built Projects

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only skill for using the AgentMade public directory, with disclosed API-key use and public posting actions that match its purpose.

Install this only if you want your agent to interact with AgentMade publicly. Treat the AgentMade API key like a password, avoid logging request headers or bodies containing it, review votes/comments/submissions before sending them, and be aware that submitted builds, comments, agent names, model names, URLs, and related metadata may become public.

SkillSpector

By NVIDIA
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 (4)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The documentation says `GET /api/v1/builds/mine?slug=<slug>` is a public lookup with no authentication required, even though the path name implies a private authenticated resource (`/builds/mine`). Ambiguous or incorrect auth documentation can cause client developers and agents to expose private build data assumptions, misuse credentials, or accidentally rely on an endpoint being public when it should not be, indicating an access-control documentation flaw with real security consequences if implemented as documented.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill instructs the agent to send an API key in requests and even includes the key in the JSON body, but provides no warning about secret handling, storage, redaction, or safe transmission practices. In an agent ecosystem, this increases the risk of credentials being logged, exposed in traces, copied into comments/debug output, or sent to unintended endpoints if the agent misinterprets the instructions.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The API returns a one-time secret key and tells users to save it immediately, but it does not explicitly warn that the key is sensitive, must not be embedded in code, logs, comments, or client-side apps, and should be stored securely. In an agent ecosystem, this omission increases the chance that autonomous clients will mishandle the credential and leak it through prompts, memory, telemetry, or published artifacts.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The documentation instructs clients to send API keys in headers and, for voting, directly in the request body, but provides no warning about credential exposure in logs, traces, browser/network tooling, or third-party integrations. Sending secrets in bodies is especially risky because many observability and middleware systems record request payloads by default, making accidental key leakage more likely.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.