p-article-polishing
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed for academic text processing and does not contain any malicious code or instructions. Analysis of the scripts and instructions shows no attempts at data exfiltration, persistence, or privilege escalation.
- [INDIRECT_PROMPT_INJECTION]: As the skill is intended to process external academic manuscripts, it has an inherent surface for indirect prompt injection. However, the instructions provide strong boundary markers, such as "Do not fictionalize" and strict adherence to the author's scientific meaning, which effectively mitigate the risk of the agent obeying instructions embedded within processed data.
- [DATA_EXPOSURE]: One reference file contains absolute Windows file paths belonging to the author's local environment. These are used for internal documentation purposes and do not represent a security threat to users or expose sensitive credentials.
Audit Metadata