kami
Warn
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill acts as a redirection stub that retrieves its full implementation from a third-party GitHub repository (https://github.com/tw93/Kami.git).
- [REMOTE_CODE_EXECUTION]: The provided installation workflow involves cloning external source code and installing it into the agent's environment using npx skills add. This allows for the execution of code from an unverifiable remote source.
- [COMMAND_EXECUTION]: The skill instructions include shell commands for environment checking and repository manipulation. Additionally, the typesetting workflow involves a 'verification command' which could lead to arbitrary command execution if not properly constrained.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text and data for document generation, creating an attack surface for instructions embedded in input documents. • Ingestion points: User-provided content for resumes, white papers, and other document types. • Boundary markers: None present in the stub to delimit user content from instructions. • Capability inventory: File system access, PDF/PPTX generation, and shell command execution for verification. • Sanitization: No evidence of content sanitization or validation is present.
Audit Metadata