huashu-proofreading
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill instructs the agent to access a hardcoded absolute file path on the local machine (
/Users/alchain/Documents/写作/写作参考/个人素材库/全部即刻动态.csv) using a grep tool. Accessing files within a specific user's home directory (e.g., Documents) is considered a sensitive operation that can lead to unintended data exposure if the execution environment matches the hardcoded path. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided articles for style adjustment, creating a surface for indirect prompt injection.
- Ingestion points: The skill reads and modifies user-supplied text articles provided during the proofreading process.
- Boundary markers: The instructions lack explicit delimiters or warnings to isolate user content from system instructions, which could allow malicious commands embedded in an article to influence agent behavior.
- Capability inventory: The skill utilizes local file system read operations (via the mentioned grep tool) and generates text output based on ingested data.
- Sanitization: There is no evidence of input validation or sanitization for the content being processed.
Audit Metadata