skills/jeremylongshore/tons-of-skills-marketplace/orchestrating-multi-agent-systems/Gen Agent Trust Hub
orchestrating-multi-agent-systems
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [SAFE]: The skill provides legitimate documentation, templates, and examples for orchestrating AI agents using the Vercel AI SDK v5.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines an architecture for processing untrusted external data.\n
- Ingestion points: User queries, support tickets, and pull request diffs handled in orchestration examples (orchestrator.ts, research-pipeline.ts, code-review-system.ts).\n
- Boundary markers: None explicitly used in the simplified code examples.\n
- Capability inventory: Tools for documentation search, simulated financial lookups (Stripe), and web searching.\n
- Sanitization: Examples utilize JSON.parse to validate structured output from LLMs.\n- [METADATA_POISONING]: Minor documentation inconsistencies were observed, such as scripts/README.md referencing .sh files while .py files are provided, and env_setup.py containing project initialization code instead of API key setup. These appear to be benign templating errors.
Audit Metadata