The old way
- Every project reinvents its own prompt patterns
- No shared vocabulary between agents and engineers
- Failures stay silent until they hit production
- Nothing is inspectable after the fact
OPEN FRAMEWORK • AI-AGNOSTIC • PORTABLE
The Standard for AI Engineering
Principles over prompts. Agent Manifesto is an open ecosystem of architectural principles, implementation kits, and test automation frameworks for building deterministic AI systems.
Model-agnostic. Provider-agnostic. Future-proof.
Claude
Codex
Gemini
Any model
Core principles
Six rules the manifesto governs by.
Load Only What You Need
Default context should contain only what the current work needs to begin correctly.
Earn Complexity
The smallest structure that solves the real current problem wins.
One Artifact, One Job
Every artifact should have one clear responsibility it can be judged against.
One Owner Per Concern
Every rule, constraint, and definition has exactly one authoritative home.
Make Behavior Explicit
Assumptions, success criteria, and stopping conditions should be stated rather than inferred.
Ask Before You Cut
Changes that can lose work or reshape authority require explicit consent.
The problem
Ad-hoc prompting doesn’t scale.
With Agent Manifesto
- One set of architectural principles, governed and versioned
- A CLI kit that adopts principles into a real repository
- Deterministic, testable agent behavior
- Every release is inspectable — version, date, diff
The ecosystem
One set of principles. Three ways to use them.
The principles
AgentManifesto
A versioned specification for building deterministic AI systems — what to do, what not to do, and why.
The CLI
Agent Manifesto Kit
Skills, agents, and bundles that adopt the manifesto’s principles into a real repository, one claude run at a time.
The companion tools
PlayForge · TaskPilot
PlayForge (Playwright + TypeScript + AI pipelines) and TaskPilot (local-first Jira for AI) — built on the same principles.