eval-chat-traces
Pass
Audited by Gen Agent Trust Hub on Apr 5, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill accesses CloudWatch and Athena logs to perform its primary function of chat trace analysis. This involves standard read-only data access within the AWS environment and does not include transmission to unauthorized external domains.
- [PROMPT_INJECTION]: The skill processes untrusted chat logs, which creates a potential surface for indirect prompt injection. Ingestion points: ChatQuery and ChatFeedback logs are retrieved from AWS CloudWatch and Athena tables (extralife_chat_logs). Boundary markers: Absent; logs are displayed in a structured Markdown table without specific delimiters or instructions to ignore embedded content. Capability inventory: The skill lacks dangerous capabilities such as file system writes, network requests to external domains, or command execution (subprocess calls/eval). Sanitization: No sanitization or filtering of the ingested log content is specified. Due to the lack of exploitable capabilities, this surface does not pose a significant risk.
Audit Metadata