Back to skill

Security audit

Agent Input Firewall

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent prompt-injection scanner, but it broadly sends potentially private agent inputs to an external service with limited disclosure and safeguards.

Review before installing. Use this only for content you are comfortable sending to the cheetahsecurity endpoint, avoid submitting secrets, private messages, authenticated tool output, source code, or personal data unless you have explicit approval, and set clear controls before using any paid x402 client flow.

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
Findings (1)

other

Warning
Location
SKILL.md:13
Finding

Unrestricted Transmission of Potentially Sensitive Content to an External Service

Content
View full analysis
"}' ``` ``` The broader instructions also identify posts, comments, direct messages, tool results, and scraped web content as eligible inputs. Lines 35–37 characterize the detection as “self-hosted” and claim “no third-party data sharing,” although the documented command sends the submitted text to an externally operated domain. ### Technical Analysis The Skill directs an agent to serialize the complete inbound text into a JSON request and transmit it to `https://x402.cheetahsecurity.de/scan`. Although TLS protects the request in transit, the remote service necessarily receives the plaintext submitted for analysis. Untrusted content is not necessarily public or non-sensitive. Direct messages, authenticated web content, and tool results can contain credentials, access tokens, personal information, internal URLs, proprietary source code, infrastructure details, or confidential conversation context. The instructions do not require: - User consent before external transmission. - Secret or personal-data detection and redaction. - Data minimization or local preprocessing. - Restrictions on direct messages, authenticated content, or tool output. - Disclosure of server-side logging, retention, deletion, or jurisdiction. - A local-only scanning alternative. Consequently, the workflow expands the data-access scope beyond what is minimally required for p ...[truncated 1711 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

External Transmission

Medium
Category
Data Exfiltration
Confidence
89% confidence
Finding

The skill instructs users to POST arbitrary untrusted text to an external internet endpoint for scanning. Even though the skill claims the service is self-hosted and does not share data with third parties, this still transmits potentially sensitive prompts, tool output, agent messages, or scraped content off the local trust boundary, creating confidentiality and compliance risk if the content contains secrets, personal data, or proprietary information.

Content

Scanner excerpt · SKILL.md (reported line 21)May include surrounding context.

How to call it (free during launch — no auth)

bash
curl -X POST https://x402.cheetahsecurity.de/scan \
  -H "Content-Type: application/json" \
  -d '{"text": "<the untrusted text your agent is about to read>"}'