kernel-audit-integration-security
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to evaluate and process untrusted external resources (scripts, iframes, API keys) provided during a development workflow. While the skill contains strong defensive instructions and enforcement rules to mitigate risks, the processing of third-party integration metadata represents a structural attack surface for indirect prompt injections.
- Ingestion points: Integration requests containing third-party URLs, script snippets, or configuration objects in SKILL.md.
- Boundary markers: The skill uses a structured 'Pre-Ship Checklist' and 'Required Review' questions to bound its analysis, though it lacks explicit input delimiters.
- Capability inventory: The skill itself does not have a command execution or network access capability; its primary output is a security decision and configuration requirements.
- Sanitization: Not applicable as the skill performs manual (LLM-driven) review rather than automated parsing.
- [SAFE]: The skill represents a defensive configuration aimed at preventing common web vulnerabilities like XSS, supply-chain attacks, and credential exposure. It explicitly prohibits risky behaviors such as using floating versions (@latest) or wildcard CSP/CORS settings.
Audit Metadata