Capacity Planning Teams AGENTS.md Template

Overview

AGENTS.md Template for capacity planning teams defines the operating manual for AI coding agents used in forecasting and capacity modeling. It supports both a single agent and robust multi-agent orchestration across planning, analysis, validation, and deployment tasks. The template makes explicit the roles, memory, source of truth, and handoff rules to ensure reliable capacity forecasts and auditable changes.

When to Use This AGENTS.md Template

• To standardize an AI-assisted capacity planning workflow across planners, analysts, and validators.
• When you need explicit agent roles, governance, and escalation paths for forecasting tasks.
• To enforce tool governance and secure data handling in production environments.
• To enable repeatable, auditable multi-agent orchestration patterns for capacity scenarios.

Copyable AGENTS.md Template

# AGENTS.md

Project role: Capacity Planning AI Ops Lead
Agent roster and responsibilities:
- Planner: defines forecasting prompts, data inputs, and planning horizon.
- Implementer: executes forecast model runs, data transformations, and rule-based adjustments.
- Reviewer: validates model outputs, checks assumptions, and approves changes.
- Researcher: gathers external signals, benchmarks, and scenario data.
- Domain Specialist: ensures domain accuracy for capacity events and constraints.

Supervisor or orchestrator behavior:
- Orchestrator coordinates all agents, sequences tasks, and maintains the single source of truth.
- Enforces timeouts, retries, and escalation when outputs are out of bounds.

Handoff rules between agents:
- Planner → Implementer on data prep and model execution tasks.
- Implementer → Reviewer on output validation and approval step.
- Reviewer → Orchestrator for final sign-off and deployment triggers.
- Orchestrator triggers retrospective review if forecast variance exceeds threshold.

Context, memory, and source-of-truth rules:
- Context is loaded from the central data lake and model catalog.
- Memory is scoped to the current planning cycle and persisted in a memory store.
- Source-of-truth: data warehouse, forecast model results, historical baselines, and official dashboards.

Tool access and permission rules:
- Tools allowed: data lake queries, model APIs, dashboard and BI tools.
- Secrets stored in vault; do not store in code.
- Only sanctioned endpoints may be called in production.

Architecture rules:
- Monorepo style with clear separation between planning, execution, and validation layers.
- No direct production data edits by agents without approval gates.

File structure rules:
- Place artifacts under capacity-planning/ with folders for data, models, and dashboards.

Data, API, or integration rules when relevant:
- Ingest data from data lake tables using defined schemas.
- Use model API endpoints for forecast runs with versioned prompts.

Validation rules:
- All outputs must pass unit tests, cross-checks against baselines, and dashboard-facing checks.
- Recompute forecasts with a new seed to verify stability.

Security rules:
- Do not leak secrets; rotate credentials; only read permissions for data.

Testing rules:
- Include unit, integration, and end-to-end tests; use synthetic scenarios for testing.

Deployment rules:
- Deploy to staging first; require approval before production rollout.

Human review and escalation rules:
- Any forecast anomaly beyond defined tolerance triggers human review and a manual rollback plan.

Failure handling and rollback rules:
- If an execution fails, revert to the last known-good forecast and alert stakeholders.

Things Agents must not do:
- Do not modify production data without approval.
- Do not bypass validation or skip tests.
- Do not share secrets in logs or outputs.

Recommended Agent Operating Model

The agent operating model defines clear roles, decision boundaries, and escalation paths to enable safe, scalable collaboration among planners, implementers, reviewers, researchers, and domain specialists. The orchestrator coordinates handoffs and ensures alignment with capacity targets and governance constraints.

Recommended Project Structure

capacity-planning-agents/
  orchestrator/
  planner/
    prompts/
    workflows/
  implementer/
    data-transformations/
    model-runs/
  reviewer/
  researcher/
  domain-specialist/
  data/
  models/
  dashboards/
  configs/
  tests/
  docs/
  scripts/

Core Operating Principles

• Single source of truth for forecasts and inputs.
• Explicit handoff boundaries with timeouts and escalation.
• Auditable decisions with versioned models and prompts.
• Least-privilege tool access and secrets management.

Agent Handoff and Collaboration Rules

• Planner hands off to Implementer for data prep and model runs.
• Implementer hands off to Reviewer for validation and acceptance testing.
• Reviewer hands off to Orchestrator for deployment gating and release notes.
• Domain Specialist and Researcher provide signals as needed and are consulted before major scenario changes.

Tool Governance and Permission Rules

• All tool use is governed by role-based access controls and audit trails.
• Secrets never appear in logs; use secure vaults and rotate tokens regularly.
• Production endpoints require approval gates and multi-person sign-offs for changes.

Code Construction Rules

• Write deterministic code with versioned prompts and data schemas.
• Validate inputs, handle edge cases, and fail fast with meaningful errors.
• Document rationale for model choices and parameter settings.

Security and Production Rules

• Follow data governance and privacy requirements for all inputs and outputs.
• Implement access controls, encryption in transit and at rest, and regular security reviews.
• Plan for rollbacks and incident response in production.

Testing Checklist

• Unit tests for data transforms and forecasting prompts.
• Integration tests for data ingestion, model runs, and outputs.
• End-to-end tests with synthetic scenarios; validate against baselines.
• Deployment tests with staging approvals.

Common Mistakes to Avoid

• Skimping on validation or relying on a single forecast without cross-checks.
• Allowing unrestricted tool access or bypassing governance gates.
• Unclear handoff boundaries leading to duplicated work or drift.

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

FAQ

What is this AGENTS.md Template for capacity planning teams?

It defines roles, governance, and collaboration rules for AI coding agents used in capacity forecasting, enabling single-agent or multi-agent orchestration.

How are agent handoffs managed in multi-agent capacity forecasting?

Planner → Implementer → Reviewer → Orchestrator; escalations trigger a manual review or rollback as needed.

What memory and source-of-truth rules govern this workflow?

Context comes from a central data lake; memory is cycle-scoped; source-of-truth includes data warehouse data, model outputs, and dashboards.

What tool access and permission controls are required?

Role-based access, vault-stored secrets, and approved endpoints; production changes require gates and approvals.

How should validation, testing, and deployment be performed?

Unit, integration, and end-to-end tests with staging gates; deployment to production only after successful reviews and approvals.