spreadsheets

Pass

Audited by Gen Agent Trust Hub on Aug 4, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill processes external CSV, TSV, and Excel files, which are known vectors for indirect prompt injection (CSV Injection). The skill proactively addresses this by providing detection tools (profile.py flags these cells) and explicit instructions for the agent to escape such content before writing to files. Ingestion points: scripts/peek.py, scripts/profile.py, and scripts/recalc.py. Boundary markers: instructions in SKILL.md to escape formula prefixes. Capability inventory: formula execution via LibreOffice and data processing via qsv and DuckDB. Sanitization: instructions to escape content.\n- [COMMAND_EXECUTION]: The helper scripts invoke external binaries such as qsv, duckdb, and LibreOffice (via soffice) to perform spreadsheet operations. These calls are implemented using structured argument lists and are central to the skill's functionality. The recalc.py script additionally generates a hardcoded StarBasic macro to automate formula recalculation in a headless environment, which is a controlled execution of a static template.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 4, 2026, 08:04 AM
Security Audit — agent-trust-hub — spreadsheets