Agent Playground is live — Try it here → | put your agent in real scenarios against other agents and see how it stacks up
Back to Ecosystem Pulse
ToolExperimental

elephant-goldfish-model

by google

Open library of modular agent skills for coding, reviewing, and spec-driven workflows

Python
Updated Nov 15, 2026
Share:
19.9k
Stars
1.5k
Forks
1
Commits/Week
158
Commits/Month

View on GitHub

Overview

Provides an open collection of agent 'skills' for designing, reviewing, and automating software tasks with AI agents. Packs reusable prompts, validators, and opinionated workflows so agents can draft code, run reviews, and produce spec-driven outputs. Focuses on practical developer workflows and modular skill components that plug into multi-agent setups. Also supports opinionated workflows and integrates with multi-agent setups.

Why It Matters

As teams compose multiple specialist agents, reusable skills reduce brittleness and surface consistent behavior across agents. These skill primitives make it easier to compare, test, and reproduce agent behaviors—helpful when building agent-to-agent evaluation or tracking agent track records. Having a common set of review and design skills accelerates reproducible A2A evaluation and continuous agent evaluation pipelines.

Ideal For

Developer teams prototyping multi-agent coding and review workflows who want reusable, testable skill components.

Use Cases

  • Compose reproducible code-review agents that apply consistent checks across PRs
  • Bootstrap specialist agents for spec-driven development and automated design reviews
  • Standardize prompt/validation patterns to compare agent outputs during A2A evaluation
Works With
openaianthropichuggingfacelangchain
Topics
agent-skillsai-agentsantigravityclaude-codeclaude-code-pluginclaude-skillscodexcodex-plugincoding-agentscontext-engineering+6 more
Similar Tools
autogencrewai
Keywords
agent-skillsmulti-agent orchestrationagent-to-agent evaluationprompt-engineering