flow-nexus-platform

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides the ability to run arbitrary code in isolated sandbox environments using the mcp__flow-nexus__sandbox_execute tool.
  • [EXTERNAL_DOWNLOADS]: Users can configure sandboxes to install external packages from official registries such as npm and pip during the setup process, and the platform utilizes Stripe for payment processing.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its processing of untrusted data across various ingestion points.
  • Ingestion points: Data entering through mcp__flow-nexus__seraphina_chat (chat messages), mcp__flow-nexus__challenge_submit (solution code), and mcp__flow-nexus__app_store_publish_app (application source code).
  • Boundary markers: None identified in the provided tool descriptions in SKILL.md.
  • Capability inventory: High-privilege actions including code execution, storage management, and application deployment.
  • Sanitization: No evidence of input sanitization or validation for processed code or assistant messages.
  • [DATA_EXFILTRATION]: The skill includes tools for managing cloud storage buckets and database subscriptions, allowing for data movement as part of standard platform operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 05:04 AM
Security Audit — agent-trust-hub — flow-nexus-platform