kelly-family-office
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from CSV files, creating a potential surface for indirect prompt injection.\n
- Ingestion points: Data enters the system via user-provided CSV files processed by
scripts/import_csv.mjs.\n - Boundary markers:
SKILL.mdexplicitly defines a read-only monitoring boundary and instructs the agent to treat holdings data as sensitive, minimizing accidental obedience to instructions embedded within imported data.\n - Capability inventory: The skill includes scripts to read local files, write to a remote database (Busabase) via ChangeRequests, and a web application (AirApp) to read and display records.\n
- Sanitization: Input data is validated against a strict schema during import, and the dashboard implementation uses HTML escaping (via the
escapeHtmlutility) when rendering data to the user interface.
Audit Metadata