OSC - Behavior-addressed software

Open Source Contracts
I got fed up with fake unit tests, building 70 assistive context helpers(LLMs, TOOLS, CLAUDE, etc…) Rewriting prompts to fit the needs of outcomes, building extensive 10k software contracts is nonsense and take more time to build out. I needed atomic units of work.
What behavior is needed to complete this task, what outcomes does it need. I started noticing a trend in what I wanted to get from models building software, and what was being produced. Claude helped form the spec, the demands were just software I wanted. Everything has a verification witness, every unit test must work on 3rdREVIEW.

This started producing working software, then I took the language contraints out of spec. Now it can make a python atomic software workflow, in ANY language it’s capable of.

The SKILL brought the entire project to work, nearly everytime, with many models, on different softwares, even in the web-chat free tier models.

I’m not a coder, but I’ve written more scripts than I can count. This made scripting LLM work much easier for my workflow.

Given three files:
{contract}.osc.md
osc.osc.md(spec)
SKILL.md (osc SKILLset)

One prompt:
Use the osc SKILL to plan and execute the osc software in this folder.

Will deliver:
Software, likely with build artifacts and test data the agent creates
A verification runner matched to this software that outputs:
A snapshot of results. Immutable(within reason) hash chained results from the runner

The 3rdREVIEW allows a second model or context window to review it as an audit. Chaining this after the build gives predictable results.

I assembled some ‘universal’ like software that could be used on any system to see the effect of the spec on outputs.

I encourage you to test it as I develop a way to unify the results snapshots on HF

Eventually this link will work once I figure out uploads on HF: