life-paths
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it is designed to ingest and process untrusted external data, including resumes, performance reviews, and financial exports, to populate the user dossier and evidence log.
- Ingestion points: Files provided by the user during the intake stage (references/intake.md) and financial data (references/financial.md).
- Boundary markers: The instructions for sub-agents include mandatory citation requirements (references/analysis-agents.md) to ensure claims are grounded in the evidence log, though it does not explicitly define structural delimiters for the untrusted content.
- Capability inventory: The skill can execute local scripts (scripts/fi_model.py), perform web searches for financial regulations, and write to the local filesystem.
- Sanitization: The skill implements a robust 'verification pass' where a fresh-context sub-agent audits the final deliverable against the evidence log to ensure no unsupported or injected claims persist (references/synthesis.md).
- [COMMAND_EXECUTION]: The skill executes a bundled Python script (scripts/fi_model.py) using a subprocess to perform Monte Carlo simulations. Additionally, the skill's logic allows the agent to generate and execute bespoke Python modules to handle complex financial modeling scenarios that the standard script cannot cover (references/financial.md).
Audit Metadata