agentic-flow-best-practices
Pass
Audited by Gen Agent Trust Hub on May 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is educational and instructional, providing patterns for building secure AI agent workflows without implementing any dangerous operations itself.
- [EXTERNAL_DOWNLOADS]: Includes references to official repositories and cookbooks from trusted vendors (Anthropic, OpenAI, Google, Microsoft, LangChain) to support implementation.
- [DATA_EXFILTRATION]: Establishes defensive policies against the storage of secrets and unredacted private information in caches or memory.
- [PROMPT_INJECTION]: Mitigates risks by mandating structured input/output schemas (Zod) and human approval gates for any action involving side effects or persistent state changes.
Audit Metadata