Workflow/Orchestration/Original by Mehansh

Loop Until Done

Autonomous self evaluation loop rewriting drafts until rubric bars clear.

What It Does

Wraps multi step tasks in an iterative cycle where the agent plans, produces, grades its output against objective criteria, identifies weaknesses, and rewrites until all requirements pass.

Architecture and Design Rationale

Initial agent completions frequently stop at minimum viable drafts. Substantial quality gains emerge from iterative evaluation against explicit rubrics.

Attribution: Original iterative workflow protocol authored by Mehansh Barthwal.

When to Activate

•Tasks requiring verified completion against objective criteria
•Repeated generation cycles where criteria can be strictly scored
•Self correcting drafting processes that operate without human intervention

When to Avoid

•Subjective design decisions or quick factual lookups

Skill Specification and Runtime Setup

Full Instruction Specification

# Loop Until Done Specification

## Overview
Wraps multi step tasks in an iterative cycle where the agent plans, produces, grades its output against objective criteria, identifies weaknesses, and rewrites until all requirements pass.

## Rationale
Initial agent completions frequently stop at minimum viable drafts. Substantial quality gains emerge from iterative evaluation against explicit rubrics.

## Core Principles
* Always verify environment constraints and dependencies before executing.
* Enforce negative constraints and reject speculative abstractions.
* Validate output against stated acceptance rubrics.

## Execution Workflow
1. Discovery: Map requirements and inspect active context.
2. Decomposition: Break down the task into discrete, verifiable phases.
3. Execution: Implement surgical changes preserving existing code architecture.
4. Verification: Test all criteria before declaring completion.