te-model-selector
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes command-line tools such as
ae-cli(e.g.,ae-cli team +list-projectsandae-cli analysis-meta event list) to retrieve metadata required for field mapping and model configuration. This is standard functional behavior for a data analysis integration and follows the principle of using provided platform tools. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and interpret natural language requirements from users to provide recommendations, which creates a surface for indirect prompt injection.
- Ingestion points: User-provided analysis requirements and project descriptions processed in Phase 1 and Phase 2.
- Boundary markers: The skill uses structured markdown tables and predefined configuration templates to delimit its output.
- Capability inventory: The skill is restricted to metadata retrieval and configuration guidance; it does not have the capability to execute investigations, write data, or perform network exfiltration. Actual execution is explicitly routed to the
ae-analysisskill. - Sanitization: Instructions strictly prohibit the fabrication of event or property names, mandating the use of
[to-replace: xxx]placeholders when real metadata is unavailable, which mitigates the risk of the agent hallucinating or being misled by malicious inputs.
Audit Metadata