# Eval Harness

> Evaluation framework for measuring agent performance

- **Type:** Skill
- **Install:** `agentstack add skill-besync-labs-antigravity-ai-kit-eval-harness`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [besync-labs](https://agentstack.voostack.com/s/besync-labs)
- **Installs:** 0
- **Category:** [Developer Tools](https://agentstack.voostack.com/c/developer-tools)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [besync-labs](https://github.com/besync-labs)
- **Source:** https://github.com/besync-labs/antigravity-ai-kit/tree/main/.agent/skills/eval-harness
- **Website:** https://besync-labs.github.io/antigravity-ai-kit/

## Install

```sh
agentstack add skill-besync-labs-antigravity-ai-kit-eval-harness
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Eval Harness Skill

> **Purpose**: Measure and improve agent performance through structured evaluation

---

## Overview

This skill provides a framework for evaluating agent performance across key dimensions.

---

## Evaluation Dimensions

### 1. Accuracy

- Did the solution work correctly?
- Were there any bugs introduced?
- Was the approach optimal?

### 2. Efficiency

- How many iterations were needed?
- Was context used efficiently?
- Were there unnecessary steps?

### 3. Alignment

- Were operating constraints followed?
- Were user preferences respected?
- Was the solution appropriate?

### 4. Quality

- Does code meet standards?
- Is it well-tested?
- Is it maintainable?

---

## Evaluation Metrics

| Metric                  | Target |
| :---------------------- | :----- |
| First-time success rate | >80%   |
| Iterations to solution  | 80%   |
| Build success           | 100%   |

---

## Report Format

```markdown
# Evaluation Report

## Session: [ID]

## Task: [Description]

### Metrics

| Dimension  | Score | Notes               |
| :--------- | :---- | :------------------ |
| Accuracy   | 9/10  | Minor fix needed    |
| Efficiency | 8/10  | 2 iterations        |
| Alignment  | 10/10 | All constraints met |
| Quality    | 9/10  | Good coverage       |

### Overall: 36/40 (90%)

### Learnings

- [What went well]
- [What could improve]
```

---

## Integration

- Run at session end for learning
- Use for continuous improvement

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [besync-labs](https://github.com/besync-labs)
- **Source:** [besync-labs/antigravity-ai-kit](https://github.com/besync-labs/antigravity-ai-kit)
- **License:** MIT
- **Homepage:** https://besync-labs.github.io/antigravity-ai-kit/

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-besync-labs-antigravity-ai-kit-eval-harness
- Seller: https://agentstack.voostack.com/s/besync-labs
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.