architect-create-tool-test
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes instructions to ground test cases in real conversation data by fetching logs from the
GET /v1/convai/conversations/{conversation_id}endpoint. - Ingestion points: Conversation data is read from the ElevenLabs API and used to populate the
chat_historyfield in test creation payloads. - Boundary markers: The skill does not specify explicit delimiters or "ignore instructions" wrappers for the ingested conversation content.
- Capability inventory: The skill uses
curlto perform network operations against theapi.elevenlabs.iodomain. - Sanitization: The instructions do not detail sanitization or filtering of the fetched conversation data before it is interpolated into the test JSON payload.
- [COMMAND_EXECUTION]: The skill utilizes
curlcommands to interact with the vendor's API for listing tools, creating test cases, and executing test suites. These commands target the official ElevenLabs API domain (api.elevenlabs.io) and are standard for this type of integration.
Audit Metadata