# harbor-framework/terminal-bench-science architecture

> Architecture diagram of the GitHub repository harbor-framework/terminal-bench-science, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-22). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/harbor-framework/terminal-bench-science
- Repository: https://github.com/harbor-framework/terminal-bench-science

## Overview

Terminal-Bench-Science is a scientific-workflow benchmark: contributors propose, implement, and review tasks, then researchers run those tasks with AI agents or oracle solutions and compare results. The repository primarily contains task packages across five scientific domains; Harbor, the leaderboard, and task dashboard are external services. The README documents the lifecycle and benchmark workflow, but sampled source does not show a central runner or results pipeline, so those internal links are intentionally omitted.

Main entry points: [`tasks`](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks) (Propose Task), [`ci_checks/rubric_review.py`](https://github.com/harbor-framework/terminal-bench-science/blob/main/ci_checks/rubric_review.py) (Task Review), [`tasks/life-sciences`](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/life-sciences) (Life Sciences), [`tasks/physical-sciences`](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/physical-sciences) (Physical Sciences)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_contribute["Task Contribution"]
  node_propose["Propose Task"]
  node_implement["Build Task"]
  node_review["Task Review<br/>[rubric_review.py]"]
end

subgraph group_benchmark["Benchmark Tasks"]
  node_life["Life Sciences"]
  node_physical["Physical Sciences"]
  node_earth["Earth Sciences"]
  node_math["Mathematical Sciences"]
  node_engineering["Engineering Sciences"]
end

subgraph group_evaluation["Benchmark Use"]
  node_harbor["Harbor Runner"]
  node_agents["AI Agents"]
  node_oracle["Oracle Solutions"]
end

subgraph group_community["Project Tracking"]
  node_leaderboard["Leaderboard"]
  node_dashboard["Task Dashboard"]
  node_releases["Tagged Releases"]
end

node_researchers(("Researchers"))
node_contributors(("Task Contributors"))

node_contributors -->|"submit proposal"| node_propose
node_propose -->|"approved proposals"| node_implement
node_implement -->|"opens pull request"| node_review
node_researchers -->|"run benchmark"| node_harbor
node_harbor -->|"evaluates"| node_agents
node_harbor -->|"runs solutions"| node_oracle
node_harbor -.->|"loads tasks"| node_life
node_harbor -.->|"loads tasks"| node_physical
node_harbor -.->|"loads tasks"| node_earth
node_harbor -.->|"loads tasks"| node_math
node_harbor -.->|"loads tasks"| node_engineering
node_researchers -->|"compare results"| node_leaderboard
node_contributors -->|"track progress"| node_dashboard
node_researchers -->|"browse coverage"| node_dashboard
node_researchers -->|"browse releases"| node_releases

click node_propose "https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks"
click node_implement "https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks"
click node_review "https://github.com/harbor-framework/terminal-bench-science/blob/main/ci_checks/rubric_review.py"
click node_life "https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/life-sciences"
click node_physical "https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/physical-sciences"
click node_earth "https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/earth-sciences"
click node_math "https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/mathematical-sciences"
click node_engineering "https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/engineering-sciences"
click node_oracle "https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_propose,node_implement,node_review toneBlue
class node_life,node_physical,node_earth,node_math,node_engineering toneAmber
class node_harbor,node_agents,node_oracle toneMint
class node_leaderboard,node_dashboard,node_releases,node_contributors toneRose
class node_researchers toneIndigo
```

## Components

### Task Contribution

- [Propose Task](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks): `tasks`
- [Build Task](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks): `tasks`
- [Task Review](https://github.com/harbor-framework/terminal-bench-science/blob/main/ci_checks/rubric_review.py): `ci_checks/rubric_review.py`

### Benchmark Tasks

- [Life Sciences](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/life-sciences): `tasks/life-sciences`
- [Physical Sciences](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/physical-sciences): `tasks/physical-sciences`
- [Earth Sciences](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/earth-sciences): `tasks/earth-sciences`
- [Mathematical Sciences](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/mathematical-sciences): `tasks/mathematical-sciences`
- [Engineering Sciences](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks/engineering-sciences): `tasks/engineering-sciences`

### Benchmark Use

- **Harbor Runner**
- **AI Agents**
- [Oracle Solutions](https://github.com/harbor-framework/terminal-bench-science/tree/main/tasks): `tasks`

### Project Tracking

- **Leaderboard**
- **Task Dashboard**
- **Tagged Releases**

### Other components

- **Researchers**
- **Task Contributors**

## Connections

- Task Contributors → Propose Task: submit proposal (evidence: `README.md`)
- Propose Task → Build Task: approved proposals (evidence: `README.md`)
- Build Task → Task Review: opens pull request (evidence: `README.md`)
- Researchers → Harbor Runner: run benchmark (evidence: `README.md`)
- Harbor Runner → AI Agents: evaluates (evidence: `README.md`)
- Harbor Runner → Oracle Solutions: runs solutions (evidence: `README.md`)
- Harbor Runner → Life Sciences: loads tasks
- Harbor Runner → Physical Sciences: loads tasks
- Harbor Runner → Earth Sciences: loads tasks
- Harbor Runner → Mathematical Sciences: loads tasks
- Harbor Runner → Engineering Sciences: loads tasks
- Researchers → Leaderboard: compare results (evidence: `README.md`)
- Task Contributors → Task Dashboard: track progress (evidence: `README.md`)
- Researchers → Task Dashboard: browse coverage (evidence: `README.md`)
- Researchers → Tagged Releases: browse releases (evidence: `README.md`)
