paper-slides

Warn

Audited by Gen Agent Trust Hub on Jul 18, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill accesses the sensitive file path ~/.claude/feishu.json to enable notification features. Accessing hidden configuration files in the user home directory is a security risk as they may contain service credentials.\n- [COMMAND_EXECUTION]: The skill dynamically generates a Python script (generate_pptx.py) and executes it to perform PowerPoint conversions. It also performs multiple shell command executions for LaTeX compilation using tools such as latexmk and pdflatex.\n- [PROMPT_INJECTION]: The skill ingests untrusted LaTeX data from the paper directory to build slide outlines and content, creating a vulnerability surface for indirect prompt injection.\n
  • Ingestion points: Research paper LaTeX source files located in paper/sections/ and paper/main.tex.\n
  • Boundary markers: None present; the agent directly processes and interpolates the extracted content into its generation logic.\n
  • Capability inventory: The skill utilizes extensive shell access (Bash), file writing permissions, and the ability to execute generated Python scripts.\n
  • Sanitization: No validation or sanitization of the input LaTeX content is performed before processing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 18, 2026, 04:52 PM
Security Audit — agent-trust-hub — paper-slides