erwa-api
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes and summarizes large volumes of untrusted, user-generated content from external Feishu groups, Telegram channels, and social media platforms.
- Ingestion points: Untrusted data is pulled into the agent's context through multiple endpoints, including
/api/v1/group_ca/user_content(raw chat messages),/api/v1/summaries(AI-generated group summaries), and/api/v1/second_kol(KOL social media posts). - Boundary markers: The skill does not provide the agent with specific delimiters or 'ignore' instructions to help it distinguish between legitimate data and potential instructions embedded within the chat logs or news content.
- Capability inventory: The agent is instructed to use this data for summarizing narratives and identifying market opportunities. Malicious actors in the monitored groups could potentially inject instructions into their messages to manipulate the agent's output or bias its recommendations.
- Sanitization: There is no documentation of input filtering or sanitization to prevent the AI from interpreting text within the ingested data as high-priority instructions.
Audit Metadata