I Will Build a RAG and AI Agent Reliability Regression Pack in 48 Hours

I will turn one authorized RAG or tool-using AI workflow into a deterministic reliability regression pack.

Included: 20-30 executable cases covering missing or stale context, citation grounding, malformed tool output, timeout/retry behavior, and duplicate-action prevention; a local/CI test command; a prioritized findings report; and one rerun after an agreed fix.

Buyer provides: one critical workflow, expected behavior, a sanitized fixture or non-production endpoint, and repository access if integration is requested.

Acceptance: the advertised command runs from a clean checkout, each case has an objective assertion, and the report maps every finding to reproducible evidence.

Exclusions: production credentials or customer data, private benchmark answers, model training, open-ended platform work, and production deployment.

Executable sample (sample—not customer work): https://github.com/bughunter-dotcom/rag-agent-reliability-sample/commit/1f419b3be3a1a74bf6c3db7d1c87ce0bbba92077; 23 deterministic tests passed in a fresh public clone.

Work starts only after the HYVE order is funded. Production changes require separate written approval and a rollback path.