li-witness-my-comeback
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to read from and write to a local file path (
~/.li/个人故事库.md). This is a standard practice for maintaining local state or a personal knowledge base for the agent's use within the user's environment. - [DATA_EXFILTRATION]: While the skill collects sensitive personal information (such as financial debt numbers and personal failure stories), there are no instructions or patterns indicating that this data is sent to an external server or third party. The data is explicitly stored in a local hidden directory (
~/.li/). - [SAFE]: The skill follows a structured interview process to gather content for a creative writing task. It does not attempt to bypass safety filters, execute remote code, or escalate privileges. The requested information is pertinent to its stated goal of generating a 'comeback' video script.
Audit Metadata