kimi-result-handling
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security vulnerabilities were detected in the instructions. The skill focuses on the safe rendering and interpretation of third-party model outputs.
- [COMMAND_EXECUTION]: The skill explicitly mandates a "Never auto-execute" policy for any code, commands, or file edits suggested by the external model (
SKILL.md,references/review-render.md). This ensures that the agent acts as a passive renderer for external suggestions, requiring explicit user approval before any system changes are made. - [PROMPT_INJECTION]: The instructions mitigate the risk of indirect prompt injection by treating external output as verbatim data to be quoted rather than instructions to be integrated into the agent's control flow. It includes specific rules to flag disagreements and maintain the original language of the output (
SKILL.md). - [DATA_EXFILTRATION]: No network communication or sensitive data harvesting patterns were identified. The skill operates on data provided by a local companion script.
Audit Metadata