cfo-advisor

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructions and associated scripts define a workflow for processing untrusted financial data, which presents a surface for indirect prompt injection attacks.\n
  • Ingestion points: The skill ingests data from two primary untrusted sources: direct user input for financial metrics (ARR, burn, etc.) defined in SKILL.md, and external JSON files (financials.json, scenarios.json) referenced in the tool commands.\n
  • Boundary markers: Absent. There are no delimiters or specific instructions for the agent to ignore potentially malicious instructions embedded within the financial data fields.\n
  • Capability inventory: The skill utilizes shell command execution to run Python scripts (scripts/burn_rate_calculator.py, scripts/financial_health_scorer.py, scripts/scenario_modeler.py) that process this external data and return results to the agent context.\n
  • Sanitization: Absent. Neither the markdown instructions nor the Python scripts perform sanitization or validation of the input data to prevent embedded text from being interpreted as instructions by the agent when it reviews the script outputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 08:23 PM