start-core/middleware

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is a technical guide that promotes secure coding practices. It includes 'CRITICAL' warnings for developers about not trusting client-sent identifiers (IDOR prevention) and properly separating validation from authorization.
  • [PROMPT_INJECTION]: No malicious instruction overrides, jailbreak attempts, or system prompt extraction patterns were detected.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: No hardcoded credentials, sensitive file access, or unauthorized network exfiltration patterns were found. Examples use placeholders like getToken().
  • [UNVERIFIABLE_DEPENDENCIES_AND_RCE]: The skill references standard, versioned packages from the TanStack ecosystem. No remote code execution or suspicious package installation commands are present.
  • [INDIRECT_PROMPT_INJECTION]: While the skill describes processing client-provided data (sendContext), it provides explicit sanitization and boundary-checking instructions (Layer 3 correctness) to mitigate injection and authorization bypass risks.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 06:26 AM
Security Audit — agent-trust-hub — start-core/middleware