afa-convert
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection through its core functionality of auditing external websites and processing analytics data.
- Ingestion points: The skill requests and processes external website URLs, GA4/Shopify analytics data, and user-provided screenshots during the '全链路转化审计' (Full Funnel Audit) and '微转化漏斗分析' (Micro-conversion Funnel Analysis) modes.
- Boundary markers: The provided instructions do not specify the use of clear delimiters or 'ignore embedded instructions' warnings when the agent parses content from the audited websites.
- Capability inventory: The skill has the capability to write and append findings to a persistent local memory file (
learnings.jsonl) following the 'Brand Memory Protocol'. - Sanitization: There is no mention of sanitization or filtering logic applied to the content extracted from external URLs before it is processed or stored in the brand memory, potentially allowing malicious content on a target site to influence future agent behavior or pollute its long-term memory.
Audit Metadata