relay
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill includes verbatim instructions intended for a successor agent (e.g., 'Do nothing without my explicit go', 'Do not read the packet back', 'stop there') which function as imperative overrides to the recipient agent's standard behavior and system prompts.- [INDIRECT_PROMPT_INJECTION]: The relay packet aggregates arbitrary data from the current session (established facts, progress, discarded attempts) into a structured prompt for the next agent, creating a vulnerability surface where malicious content in these fields could be interpreted as instructions by the successor model.
Audit Metadata