alibabacloud-flink-ai-ops
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the Flink development console and user-supplied messages through the
chat_ai_agenttool. This creates a surface for indirect prompt injection if an attacker can manipulate resource names or logs in the cloud console. However, the skill implements Human-in-the-Loop (HITL) for all write operations, requiring explicit user approval before any changes are committed. - [DYNAMIC_EXECUTION]: The
sse.pymodule employs runtime function patching to harden the underlying SDK's SSE (Server-Sent Events) parser. It usesFunctionTypeto inject a custom parser that enforces an 8 MiB limit on event frames and improves UTF-8 validation. This is a defensive implementation to prevent resource exhaustion and ensure data integrity during streaming conversations. - [DATA_EXPOSURE]: The skill handles sensitive Alibaba Cloud credentials (AccessKey ID/Secret). It includes a comprehensive redaction mechanism in
redaction.pythat identifies and masks these credentials in all tool outputs and event streams before they are presented to the user or logs.
Audit Metadata