output-formatter

Pass

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: LOWPROMPT_INJECTION
Full Analysis
  • [Indirect Prompt Injection] (LOW): The skill's primary purpose is to ingest and format data from other agents, which constitutes an indirect prompt injection surface.
  • Ingestion points: Processes '智能体输出结果' (Agent output results) which are potentially untrusted external inputs.
  • Boundary markers: No specific delimiters or 'ignore' instructions are defined to separate the data from the agent's formatting instructions.
  • Capability inventory: The skill has no allowed tools (allowed-tools: []), meaning it cannot perform network requests, file operations, or command execution, which limits the potential impact of an injection.
  • Sanitization: No input validation or sanitization is mentioned beyond requiring 'parseable formats'.
  • [NO_CODE] (SAFE): The skill contains no scripts, binary files, or tool definitions. It functions entirely through natural language instructions to the model.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 16, 2026, 11:06 AM