Project Documentation

Terminus−2.0

The complete coding-expert handbook for designing, building, validating and defending Terminal-Bench tasks - presented by Cognyzer.

Read the docs →

6

Sections

35

Chapters

9

Categories

5

Subtypes

Documentation

Six sections · 35 chapters · designed for Coding Experts and Reviewers.

01

Getting Started

Onboarding, environment setup and the CLI surface

01Welcome 02Quick Start 04Cli User Guide 05Edition1 To Edition2 Changes

02

Submission Process

From skeleton to reviewer hand-off

06Platform Submission 07Submission Checklist 08After Submission

03

Understanding Tasks

Anatomy, taxonomy, rubrics and difficulty

09What Makes A Good Task 10Task Components 11Task Requirements 12Task Taxonomy 13Diversity Requirements 14Milestones 15Task Subtypes 16Rubrics 17Prompt Styling 18Difficulty Guidelines 19Example Tasks

04

Creating Tasks

Hands-on guides for environments, oracles and tests

20Creating a Task 21Running Your Task 22Creating Docker Environment 23Writing Oracle Solution 24Writing Tests

05

Testing & Validation

Oracle, NOP, real agents, CI and LLM-as-Judge

25Oracle Agent 26Nop Agent 27Running Real Agents 28Ci Feedback Training 29Ci Checks Reference 30Llmaj Checks Reference 31Agent Review Reference

06

Reviewing Tasks

Reviewer playbook, common errors and defending submissions

32Review Guidelines 33Reviewer Checklist 34Reviewer Training 35Common Errors 36Defending Your Submission