Back to skill

Security audit

ClaimCheck

Security checks for vulnerabilities and agentic risk

Overview

ClaimCheck has a coherent privacy purpose, but its local-only promises conflict with proxy instructions that forward requests and API credentials to an external provider.

Review the conflicting privacy promises before installing. Proxy mode sends sanitized requests and your API authentication to the selected provider; missed sensitive values may also leave your machine. Choose your intended provider and its matching credentials, inspect sanitized output, and verify the external release before enabling the Windows startup task. The supplied static scan reported no findings, but that does not resolve the documented privacy ambiguity.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

The documented upstream proxy contradicts the manifest's unconditional 'zero network calls' claim.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The unconditional claims that nothing leaves the machine and that there are zero network calls conflict with the documented proxy forwarding requests to an external LLM provider. In a privacy-focused skill, this ambiguity can lead users to approve processing under an incorrect local-only assumption; sanitized content and any missed sensitive values can still leave the machine.

Content

No source excerpt is available for this finding.

Proxy instructions omit a clear warning that requests and API credentials leave the machine for the configured upstream, contradicting the local-only assurances.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The instructions say authentication passes through untouched while providing a concrete third-party upstream, but do not explicitly require that the client's API key belong to that upstream. Despite the visible proxy description, the surrounding local-only assurances can obscure this trust boundary and encourage forwarding credentials or residual sensitive data to an unintended provider. This overlaps with the misleading assurance finding rather than establishing a separate malicious exfiltration mechanism.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 77)May include surrounding context.

Any OS (manual): run the binary as the proxy:

text
ClaimCheck --proxy --upstream https://api.z.ai/api/coding/paas/v4/ --port 8766

Point your client's Base URL at the proxy — it mirrors the upstream path: