founder-cfo

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local Python script, unit_economics.py, to process business data and generate financial reports.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied JSON data and renders it into a Markdown report which is subsequently summarized by the AI agent, creating a surface for instruction injection via data fields.
  • Ingestion points: Data is ingested from a user-specified path or the default founder/numbers.json via the unit_economics.py script.
  • Boundary markers: No explicit delimiters or instructions are used to separate the tool's output from potential malicious instructions embedded in the data.
  • Capability inventory: The Python script performs file operations and generates text; the agent reads and summarizes this text to produce a CFO note.
  • Sanitization: Input strings like business names and cost labels are not sanitized or escaped before being placed into the Markdown report.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 09:32 AM