fengtang-perspective
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements an 'Agentic Protocol' that instructs the agent to use WebSearch tools to gather facts before responding. This creates an attack surface where untrusted data from external websites could enter the agent's context. However, the skill lacks high-risk capabilities such as file system modification, command execution, or credential access, significantly limiting the potential impact of such an injection.
- Ingestion points: WebSearch tool outputs triggered by 'Step 2: 冯唐式研究' in
SKILL.md. - Boundary markers: Absent; the instructions do not explicitly tell the agent to ignore instructions embedded in the search results.
- Capability inventory: No risky capabilities detected; the agent's actions are limited to generating natural language responses.
- Sanitization: Absent; search results are interpolated directly into the response generation process.
Audit Metadata