metro-ai-app-builder
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The orchestrator fetches the live skill catalog and technical configuration (
skills-config.json) from the official open-edge-platform GitHub repository usingcurlas part of its discovery process. - [REMOTE_CODE_EXECUTION]: The skill dynamically installs and invokes delegate skills using the
npx skillsCLI tool, pulling from the vendor's own curated skills repository upon user approval. - [COMMAND_EXECUTION]: System tools such as
bash,git, andghare utilized to perform environment checks, discover remote resources, and coordinate the deployment of Edge AI applications. - [DYNAMIC_EXECUTION]: The skill's primary function is to identify, load, and delegate tasks to other skills at runtime, which is standard behavior for an orchestration-focused agent.
- [INDIRECT_PROMPT_INJECTION]: During the discovery phase, the agent ingests remote markdown instructions (
SKILL.md) and metadata from the catalog repository. While this represents a data ingestion surface, the content is sourced exclusively from the vendor's trusted infrastructure.
Audit Metadata