Flownix
Sections
On this page

What a Harness is

A Harness defines a repeatable way for agents and people to investigate a question, record a decision, and turn it into the next plan of work.

A Harness is neither a chat nor a prompt collection. It is a versioned process: roles, a step graph, access rules, limits, expected inputs, and outputs. The same Harness can be safely repeated for different work.

  1. 01Version
  2. 02Run
  3. 03Analysis
  4. 04Review
  5. 05Decision
  6. 06Approval
  7. 07Task plan

When to use it

Architecture choices

Independent experts propose options, test assumptions, and leave an explainable decision.

Risk and security

Dedicated roles look for threats and objections; risky actions pass through a human confirmation.

Decomposition

After a decision, the system prepares an Flownix task draft for a person to review before applying it.

The core principle

Instead of retaining hidden reasoning, a Harness records inspectable work artifacts: proposals, claims, evidence, critiques, votes, and the final decision. This makes it possible to see what was chosen and why.