round-table
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements a multi-agent workflow where user-supplied topics and options are processed by champion subagents, whose outputs are then combined into a transcript for a synthesizer agent.
- Ingestion points: User input for topics and options is integrated into the agent context in SKILL.md (Step 1).
- Boundary markers: The instructions in SKILL.md (Step 4) do not specify the use of robust delimiters or 'ignore embedded instructions' warnings when presenting the collected debate transcript to the synthesizer subagent.
- Capability inventory: Champion subagents are instructed to perform external research (Step 2) and exchange direct messages, which could introduce untrusted or adversarial content into the debate transcript processed by the final agent.
- Sanitization: The workflow does not include a step to sanitize or validate the content generated by the champion agents before it is analyzed by the high-capability synthesizer.
Audit Metadata