fsi-strip-profile

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses soffice (LibreOffice) and pdftoppm (Poppler) in a headless mode to convert PowerPoint files into PDF and then into JPEG images. This is part of a mandatory visual review process to check for text overlap and layout integrity.
  • [EXTERNAL_DOWNLOADS]: The skill references data collection from various financial platforms including BamSEC, SEC EDGAR, Bloomberg, FactSet, and CapIQ. These are well-known, trusted professional sources for financial information.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it incorporates data from untrusted external sources (public news, company filings) into the agent's reasoning and slide generation workflow.
  • Ingestion points: Company news, SEC filings (Item 1, MD&A), and press releases from the last 90 days are ingested from the web.
  • Boundary markers: There are no explicit instructions for the agent to treat the research data as untrusted or to ignore embedded instructions.
  • Capability inventory: The agent can execute shell commands (soffice, pdftoppm) and generate code for the PptxGenJS library.
  • Sanitization: The skill lacks data sanitization or validation logic for the retrieved financial information before using it in slide creation.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 03:21 PM
Security Audit — agent-trust-hub — fsi-strip-profile