agent-simplification
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of existing agent configurations and conversation transcripts to generate new prompt logic and test suites.
- Ingestion points: Agent JSON and conversation histories are retrieved via GET requests to ElevenLabs API endpoints.
- Boundary markers: The skill does not define specific delimiters to isolate potentially malicious instructions within existing agent prompts during the simplification process.
- Capability inventory: The agent has permissions to modify agent configurations and create test objects using the ElevenLabs platform.
- Sanitization: The instructions explicitly mandate the removal of PII (names, numbers, and specific dates) from conversation data before it is reused in test scenarios.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill utilizes ElevenLabs API endpoints for data retrieval and modification. These operations are consistent with the skill's stated purpose and target the author's own infrastructure, posing no risk of unauthorized exfiltration.
Audit Metadata