carocut-shared-schema

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: Indirect prompt injection surface via raw user request propagation.\n
  • Ingestion points: The file references/dispatch-context.md defines an amendment.user_request field which is intended to capture "用户原话" (user's original words) and requires the system to "原样传递" (pass as-is) this content to sub-agents.\n
  • Boundary markers: The schema does not specify the use of delimiters or specific "ignore" instructions to prevent downstream agents from obeying malicious commands embedded in these user requests.\n
  • Capability inventory: As detailed in references/resource-types.md, sub-agents (Media, Builder) have capabilities including stock media retrieval, AI generation of images and sprites, TTS generation, and implementation of React components.\n
  • Sanitization: The provided schema lacks any requirement for sanitizing or validating external user input before it is consumed by the multi-agent pipeline.\n- [NO_CODE]: The skill consists entirely of Markdown-based schema definitions and instructions; it does not contain any executable scripts or binary files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 07:14 PM
Security Audit — agent-trust-hub — carocut-shared-schema