Agents | 2ndSys
Agents
Operational contract for AI participants in the engineering coordination system.
AI Engineering

Agents

AI engineering agents are participants within the Align engineering coordination system.

They are expected to preserve the same architectural integrity expected of human engineers.

Purpose

Operational Contract

AI engineering agents shall preserve the architectural integrity of the Align platform.

Before proposing, modifying, or generating code, agents shall understand the platform's governing theory, architecture, and engineering principles.

Required Reading

Read before making architectural decisions

  1. Canonical Foundation
  2. Canonical Platform Architecture
  3. Engineering Principles

These documents govern every implementation.

Workflow

Engineering Workflow

  1. Identify the business capability being introduced.
  2. Determine whether it can be expressed through Diagnostic Definitions.
  3. Preserve Platform Architecture.
  4. Evaluate the implementation against the Five Engineering Principles.
  5. Explain any architectural tradeoffs.
  6. Only then generate implementation.
Review

Required Review

Every pull request should explain:

  • What business capability is being introduced?
  • Which runtime responsibility changes?
  • Which Engineering Principles are strengthened?
  • Which principles are weakened?
  • Why was this tradeoff accepted?
Boundaries

Prohibited Behaviors

AI engineering agents shall not:

  • Duplicate runtime responsibilities.
  • Embed business policy within runtime implementation.
  • Introduce hidden constraints.
  • Reduce State Traceability.
  • Allow Decision Authority to drift.
  • Increase runtime complexity when Diagnostic Definitions can express the same capability.
Preferences

Preferred Behaviors

  • Declarative knowledge.
  • Stable runtime responsibilities.
  • Explainable execution.
  • Explicit authority.
  • Localized failures.
  • Evolution through knowledge.
Final Check

Architectural Review

Every significant implementation should answer:

  • Can this capability be expressed through knowledge?
  • Does this preserve Platform Architecture?
  • Does this preserve the Five Engineering Principles?
  • Would this implementation still make sense two years from now?