academic-latex-layout
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The workflow defined in
references/pdf-preview-loop.mddirects the agent to "Discover the build command from project files or user instructions" and "Prefer existing commands over inventing new ones." This instruction allows the execution of untrusted shell commands defined within the project's own build scripts, such as aMakefileorlatexmkrcconfiguration file. - [PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection as it processes untrusted LaTeX projects and build configurations without adequate sanitization.
- Ingestion points: LaTeX source files (
.tex), bibliography files, and project build scripts (Makefile,latexmkrc). - Boundary markers: The instructions lack explicit requirements for using delimiters or boundary markers to prevent the agent from following instructions embedded within the project files.
- Capability inventory: The skill is granted capabilities for file system read/write and shell command execution through various LaTeX build tools.
- Sanitization: There is no specified mechanism for sanitizing or validating build commands discovered in the project environment before they are executed.
Audit Metadata