ClawHub
Back to skill

Security audit

codex-supergraph

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only Codex GraphQL helper that uses API credentials and includes disclosed admin-style templates, but I found no hidden execution, exfiltration, or destructive behavior.

Install only if you intend the agent to help build Codex GraphQL/API requests. Keep the long-lived CODEX_API_KEY protected, do not allow token creation/deletion or webhook creation/deletion unless you explicitly requested it, and review any callback URL or returned token before use.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
91% confidence
Finding
The skill is described as an analytics/query-building aid, but the template document also teaches API token creation and deletion. That expands the capability surface from read-only data access into credential lifecycle management, which can enable unauthorized token minting, privilege extension, or accidental credential disruption if an agent follows these templates in response to user prompts.

Description-Behavior Mismatch

Medium
Confidence
94% confidence
Finding
Webhook creation and deletion are side-effecting remote operations that go beyond the stated read-only analytics purpose of the skill. Including operational automation templates in a query helper increases the risk that an agent provisions callbacks to attacker-controlled endpoints, creates persistent alerts, or deletes existing webhook infrastructure without the user understanding the consequences.

Context-Inappropriate Capability

Medium
Confidence
88% confidence
Finding
Documenting API token lifecycle management inside an analytics-focused skill exposes privileged capabilities that are not justified by the manifest. In an agent setting, capability overexposure is dangerous because prompt injection or user confusion can steer the agent from harmless data retrieval into issuing or revoking credentials.

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
The webhook provisioning examples introduce infrastructure-changing capabilities that are not justified by a manifest framed around analytics and GraphQL queries. In practice, this can let an agent configure outbound data flows or persistent automations to third-party endpoints, materially increasing abuse and data-exfiltration risk.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.