kelly-legal-contracts
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of legal contracts and intake JSON payloads.
- Ingestion points:
scripts/ingest_contracts.mjsreads contract facts and clause text from local JSON files. - Boundary markers: Present in
SKILL.md. The skill explicitly identifies a 'Boundary' section and uses 'risk:gated-write' tags to enforce human oversight. - Capability inventory: The skill includes scripts for file writing (
scripts/export_issues.mjs) and network operations to the Busabase platform (scripts/execute_decisions.mjs). - Sanitization: The application code uses HTML escaping (
escapeHtml) in the UI components (app.js,review-views.js) to prevent cross-site scripting from contract content. - [EXTERNAL_DOWNLOADS]: The skill relies on standard dependencies from the NPM registry.
- Evidence: Downloads
busabase-sdk,hono, and@hono/node-serverfrom npmjs.com. - [COMMAND_EXECUTION]: The skill provides several management scripts intended to be executed by the agent or user for data ingestion, risk checking, and exporting.
- Evidence: Scripts like
scripts/run_checks.mjsandscripts/export_issues.mjsperform local processing and filesystem updates as part of the legal desk's core functionality.
Audit Metadata