Snowflake Production Architecture AGENTS.md Template

Overview

A copyable AGENTS.md template that governs an agent workflow for Snowflake production architecture. It supports both single agent and multi-agent orchestration with clear sources of truth, escalation paths, and governance. This page provides a ready to paste AGENTS.md block that you can adapt for your Snowflake production environment.

Direct answer: This AGENTS.md template defines roles, handoffs, memory, and governance so AI coding agents can operate reliably in a Snowflake production architecture while enabling safe multi-agent collaboration.

When to Use This AGENTS.md Template

• You are building or sustaining Snowflake production pipelines and require formal operating context for agents.
• You need explicit handoff rules between planner, implementer, reviewer, tester, researcher, and domain specialist in a Snowflake data stack.
• You want centralized memory and source of truth for agent decisions, results, and audit trails.
• You require tool governance, secrets handling, and production safe-guards for Snowflake objects and data.
• You expect multi-agent orchestration with clear escalation and rollback paths.

Copyable AGENTS.md Template

# AGENTS.md

Project: Snowflake Production Architecture Automation

Agent roster and responsibilities
- Planner: designs tasks, defines constraints, creates the execution plan for Snowflake objects and pipelines
- Implementer: implements SQL objects, tasks, procedures, and data pipelines in Snowflake
- Reviewer: validates outputs, data quality, and schema changes against acceptance criteria
- Tester: runs unit and integration tests for Snowflake objects and pipelines
- Researcher: sources authoritative guidance, data sources, and governance requirements
- Domain Specialist: ensures Snowflake security, access controls, and network policies
- Incident Commander: oversees incident response and rollback coordination

Supervisor or orchestrator behavior
- The orchestrator coordinates tasks across agents, enforces SLAs, and triggers handoffs
- Maintains a centralized memory store for context, decisions, and results
- Enforces source of truth and auditability for Snowflake changes

Handoff rules between agents
- Planner to Implementer: hand off execution plan with constraints and acceptance criteria
- Implementer to Reviewer: hand off implemented changes with validation results
- Reviewer to Tester: hand off tested results and verification artifacts
- Researcher to Domain Specialist: request guidance for approvals or policy checks as needed
- Domain Specialist to Incident Commander: escalate security or governance issues as required

Context memory and source of truth rules
- All decisions and outputs are recorded in a central Snowflake context table and audit log
- Use a versioned change log for all DDL, DML, and pipeline changes
- Source of truth includes objectives, constraints, acceptance criteria, and test results

Tool access and permission rules
- Agents operate under least privilege in Snowflake with role based access
- Secrets and credentials stored in a secure vault; agents never hard code secrets
- External API calls and data access go through approved connectors with governance checks

Architecture rules
- Multi-layer Snowflake architecture with staging, core, and mart layers
- Separate warehouses for planning, development, and production work
- Idempotent operations and explicit rollback paths
- Logging and observability for all agent actions

File structure rules
- Keep all agent related artifacts under a dedicated repository folder
- Do not place production credentials in code or docs
- Use clear naming for tasks, pipelines, and objects

Data, API, or integration rules when relevant
- All data changes follow the data governance policy and have approvals
- API integrations must be versioned and audited
- Use test data when validating in non-prod environments

Validation rules
- Each change must pass unit and integration tests
- Data quality checks must meet predefined thresholds
- Metadata updates must be reflected in the context store

Security rules
- Encrypt data at rest and in transit
- Enforce least privilege, rotation of credentials, and access audits
- Require approvals for production changes

Testing rules
- Unit tests for individual Snowflake objects
- Integration tests for pipelines and data flows
- End-to-end tests for critical production scenarios

Deployment rules
- Use PRs with reviews before merging to main
- Deploy changes through staged environments with guardrails
- Validate in production-like environment before cutover

Human review and escalation rules
- Complex or high-risk changes require domain specialist and security review
- If tests or acceptance criteria fail, escalate to human review for remediation

Failure handling and rollback rules
- If a failure occurs, revert to the last known good state in Snowflake
- Preserve audit logs and notify stakeholders

Things Agents must not do
- Do not bypass the orchestrator or governance gates
- Do not run production changes without approved plans and tests
- Do not exfiltrate data or reveal secrets in any channel
- Do not introduce uncontrolled side effects or architecture drift

Recommended Agent Operating Model

The recommended operating model defines agent roles, decision boundaries, and escalation paths for Snowflake production architecture. The Planner designs the plan, the Implementer executes, the Reviewer validates, the Tester confirms, and the Domain Specialist ensures security and policy alignment. Handoffs are gated by validated artifacts. Escalation paths exist for security or data risk events, with the Incident Commander coordinating remediation.

Recommended Project Structure

snowflake-prod-arch/
├── agents/
│   ├── planner/
│   ├── implementer/
│   ├── reviewer/
│   ├── tester/
│   ├── researcher/
│   └── domain-specialist/
├── orchestrator/
├── data/
│   ├── sources/
│   └── marts/
├── pipelines/
├── integrations/
├── configs/
│   └── secrets/
├── docs/
└── tests/

Core Operating Principles

• Clear ownership and auditable changes for every Snowflake artifact
• Idempotent tasks with explicit rollback paths
• Centralized context and source of truth for all decisions
• Least privilege and secure secrets management
• Formal handoffs with validated artifacts

Agent Handoff and Collaboration Rules

• Planner to Implementer: provide plan document, constraints, and success criteria
• Implementer to Reviewer: provide implemented objects, test results, and validation asserts
• Reviewer to Tester: provide acceptance artifacts and pass/fail criteria
• Researcher to Domain Specialist: request policy and governance guidance as needed
• Domain Specialist to Incident Commander: escalate security or architecture risks

Tool Governance and Permission Rules

• Only approved connectors for external data sources
• Secrets stored in secure vaults; rotate per policy
• Production changes require planning and tests prior to deployment
• All API calls and data access are auditable
• Live Snowflake objects are immutable without approval gates

Code Construction Rules

• Write modular SQL and scripts with clear inputs and outputs
• Document a single source of truth for each artifact
• Use parameterization to avoid hard coding values
• Keep changes small and reviewable
• Validate inputs and handle edge cases

Security and Production Rules

• Enforce role based access control in Snowflake
• Encrypt data in transit and at rest; manage keys securely
• Require approvals for production deployments
• Monitor and alert on anomalous activity

Testing Checklist

• Unit tests for each Snowflake object
• Integration tests for pipelines and data flows
• End-to-end tests for critical production scenarios
• Rollback tests and disaster recovery drills

Common Mistakes to Avoid

• Skipping the planner review or bypassing the orchestrator
• Uncontrolled architecture drift or unapproved production changes
• Neglecting secrets management and access control
• Inadequate testing or missing audit trails

Related implementation resources: AI Use Case for Sales Pipeline Reviews and Deal Risk Scoring and AI Use Case for Corporate Event Managers Using Slack To Orchestrate Day-Of Venue Tasks Across Multi-Department Teams.

FAQ

What is the purpose of this AGENTS.md Template for Snowflake production architecture?

This template provides a structured operating manual for AI coding agents to manage Snowflake production architectures, enabling clear roles, handoffs, and governance.

Who should use this template and how should it be adapted?

Engineers, data architects, and product teams can adapt the template to their Snowflake environment by aligning roles to the actual team and by tailoring the acceptance criteria and data sources.

How are handoffs governed between planner implementer reviewer tester researcher domain specialist?

Handoffs are gated by validated artifacts such as plans, implemented outputs, test results, and policy guidance. Each handoff requires artifact validation before proceeding.

What are the security considerations for Snowflake pipelines?

Apply least privilege, secrets management, encryption, access audits, and forced approvals for production changes. Monitor for anomalous activity and enforce governance policies.

How should testing and deployment be performed?

Use unit and integration tests for each artifact, conduct end-to-end tests for critical flows, and deploy through staged environments with approvals and rollbacks available.

What happens if a production change fails?

The system rolls back to the last known good state, preserves an audit trail, and notifies stakeholders for remediation.