workpaper-standards

Pass

Audited by Gen Agent Trust Hub on Aug 15, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes external, untrusted financial data such as bank statements, invoices, and general ledger exports, which provides a surface for indirect prompt injection attacks.
  • Ingestion points: Raw source data is ingested into the Source tab (defined in SKILL.md), and OCR-extracted fields are used for calculations and memos (described in references/evidence-table-template.md).
  • Boundary markers: The instructions do not define delimiters or boundary markers to segregate untrusted source data from agent instructions.
  • Capability inventory: The skill generates structured Excel workbooks, CSV files, and journal entries based on these inputs.
  • Sanitization: There is no requirement defined to sanitize or escape text from source documents before they are used in memos or calculations, although the skill does require flagging low-confidence items for manual review.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 15, 2026, 04:09 AM
Security Audit — agent-trust-hub — workpaper-standards