research-presentation

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes local system tools such as LibreOffice (soffice) and Poppler (pdftoppm) via subprocess.run in scripts/render_slides.py. These calls are used for converting presentations to PDF and PNG formats. The implementation is secure as it uses argument lists to prevent shell command injection.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from untrusted external sources, specifically PDF research papers and manuscripts (ingestion points: SKILL.md, references/paper-extraction.md). It possesses capabilities to execute rendering scripts and write files to the local environment. To mitigate potential injection risks from external data, the skill implements a strict evidence ledger and grounding contract (references/paper-extraction.md) and a structural audit process (scripts/audit_pptx.py) that verifies all output is anchored to the source material.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 02:29 AM
Security Audit — agent-trust-hub — research-presentation