li-story-vlog 故事事件写作
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill functions entirely as a creative writing guide for producing restrained, observation-based vlog scripts. It does not utilize any tools, access sensitive files, or perform network operations. The interaction is limited to standard text generation within the agent's safe environment.
- [NO_CODE]: The skill package contains only Markdown instructions and configuration YAML. There are no associated scripts (Python, Node.js, Shell), which eliminates the risk of code injection, dependency attacks, or runtime execution vulnerabilities.
- [INDIRECT_PROMPT_INJECTION]: An assessment of the data ingestion surface was conducted:
- Ingestion points: The skill collects raw story materials from the user, including details about people, events, and commercial product points (SKILL.md, Step 1).
- Boundary markers: The instructions lack specific delimiters or isolation warnings for external data, relying on the agent's internal context management.
- Capability inventory: The skill possesses no exploitable capabilities; it cannot write files, execute shell commands, or make network calls.
- Sanitization: No specific sanitization or filtering logic is applied to the input.
- Conclusion: While the skill processes untrusted user data, the lack of dangerous tools or system access means the attack surface is not exploitable.
Audit Metadata