data-persistence-contract

What problem does it solve?

This Skill prevents AgentHub’s history, runs, artifacts, and tool-calls from being lost or only stored in ephemeral memory or logs, by defining an explicit persistence contract for a multi-agent, single/group conversation system.

Core Features & Use Cases

• Defines facts vs. non-facts: establishes what must be durably stored in the database versus what must remain temporary (SSE buffers, runtime caches, logs).
• Standardizes entity relationships: specifies how Conversation, Message, Run, RunStep, AgentTask, ToolCall, and Artifact must relate for traceability across multi-agent runs.
• Sets schema, indexing, and JSON rules: mandates snake_case DB fields, camelCase API projections, index plans, structured JSON constraints, soft-delete/archival boundaries, and MySQL/PostgreSQL evolution boundaries.
• Enforces migration and safety: requires versioned migrations with expand/migrate/contract patterns and strict data security to avoid storing secrets, raw prompts, tokens, and permanent public URLs.
• Enables contract testing: lists contract-test expectations to verify schema creation, core relationships, JSON validity, soft-delete integrity, and trace-based troubleshooting.

Quick Start

Use the data-persistence-contract Skill as the single source of truth for designing or updating AgentHub’s MySQL schema and migrations, ensuring every run artifact and tool call is persistently linked and auditable.