skills/x-cmd/skill/minimax-pdf/Gen Agent Trust Hub

minimax-pdf

Pass

Audited by Gen Agent Trust Hub on Sep 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Several scripts, including fill_inspect.py, fill_write.py, merge.py, reformat_parse.py, and render_body.py, contain an internal ensure_deps function that uses subprocess.check_call to automatically install missing Python dependencies from the official package registry. The main make.sh script also executes shell commands to install both Python and Node.js dependencies.
  • [EXTERNAL_DOWNLOADS]: The skill is configured to fetch typography styles from Google Fonts via CSS @import rules in the cover rendering stage. Additionally, it may download the Chromium browser using the Playwright utility during its initial setup or execution to facilitate PDF rendering.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the REFORMAT route, which parses Markdown, plain text, and existing PDF files. This creates a surface for indirect prompt injection where a maliciously crafted document could attempt to influence the agent's behavior during the parsing process.
  • Ingestion points: The reformat_parse.py script reads content from .md, .txt, and .pdf files provided by the user.
  • Boundary markers: The parsing logic does not implement explicit delimiters or warnings to ignore instructions embedded within the processed document content.
  • Capability inventory: The skill has the capability to write files to the local system and execute package management commands via subprocesses.
  • Sanitization: The reformat_parse.py script uses regular expression matching to extract and transform specific markdown elements into ReportLab XML, providing basic structural validation but no semantic filtering of the text content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 26, 2026, 12:02 PM