paper-2-web
Warn
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill directs the execution environment to clone an external codebase from a third-party GitHub repository (
https://github.com/YuhangChen1/Paper2All.git) and run a package installation command (pip install -r requirements.txt). This can lead to the installation and runtime execution of unverified third-party dependencies. - [INDIRECT_PROMPT_INJECTION]: The skill functions as an automated pipeline that ingests and processes untrusted user-supplied documents (academic papers in PDF or LaTeX formats) through an LLM content extraction process while holding extensive execution capabilities.
- Ingestion points: External research files (
main.tex,paper.pdf) are loaded into the workspace as described inSKILL.mdandreferences/paper2web.md. - Boundary markers: No specific delimiters or safety instructions are specified to segregate the text of the paper from the operational prompt context.
- Capability inventory: The skill possesses permission to use
Bash,Read,Write, andEdittools, allowing it to invoke shell commands and write local files during pipeline execution. - Sanitization: There is no evidence of preprocessing filters or text sanitization to strip potential malicious prompt instructions embedded inside document sections or comments.
Audit Metadata