onchain-automation
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes architectures that ingest data from external sources, such as Turbo pipelines and HTTP webhooks, which then trigger onchain actions like
writeContract. This creates an inherent surface for indirect prompt injection where malicious onchain data or webhook payloads could attempt to influence the logic of the automation. The skill mitigates this risk by explicitly instructing the agent and user to treat incoming data as untrusted and perform decoding and validation before execution. - [SAFE]: The skill's instructions regarding communication with the user (e.g., "Do NOT tell the user...") are designed to ensure technical accuracy about the platform's execution runtime (Compose) and do not represent an attempt to conceal malicious activity or bypass safety guidelines.
Audit Metadata