minimax-pdf
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill's scripts and unified CLI (
make.sh) automatically fetch and install required libraries from well-known registries, including Python packages from PyPI and the Playwright Node.js library from npm. It also fetches fonts from Google Fonts at render time. - [COMMAND_EXECUTION]: Multiple Python scripts (
merge.py,render_body.py,reformat_parse.py, etc.) utilizesubprocess.check_callto ensure necessary dependencies are present by invoking the system's package manager. The shell commands are limited to legitimate environment setup and internal orchestration. - [INDIRECT_PROMPT_INJECTION]: The skill accepts untrusted external files (Markdown, PDF, Text) as input for its reformatting pipeline. This represents a potential surface for indirect prompt injection.
- Ingestion points: External documents are read by
reformat_parse.pyto generate the internalcontent.jsonstructure. - Boundary markers: There are no explicit delimiters used to isolate processed text from the agent's instructions, though the content is structured into specific block types.
- Capability inventory: The skill can write PDF files to the local file system and execute dependency installation commands.
- Sanitization: The skill parses structural elements (headings, tables, lists) but does not specifically filter for natural language instructions embedded within the text.
Audit Metadata