{"slug": "elephant-goldfish-model-skills-for-designing-and-reviewing-software-with-agents", "title": "Elephant-goldfish-model: skills for designing and reviewing software with agents", "summary": "Google released Elephant-Goldfish-Model (EGM), an open-source plugin that lets engineers draft software specifications with a deep-context \"Elephant\" agent and stress-test them with zero-context \"Goldfish\" subagent reviewers before writing code, and thousands of Google engineers already use it. EGM installs via plugin commands for the agy (Antigravity), Claude Code and Codex coding agents, and its skills — egm-brainstorm, egm-draft, egm-review and egm-resolve — produce concepts_brief.md and design_draft.md artifacts, with reviews halting if the agent cannot create fresh subagent contexts. The plugin is aimed at features needing design decisions before implementation, not small bug fixes, and Google notes a review does not replace tests or human judgment.", "body_md": "Design with an elephant. Review with a goldfish.\n\n[Install](#install)  · \n  [Quickstart](#try-your-first-design)  · \n  [Documentation](https://github.com/google/elephant-goldfish-model/blob/main/docs/getting_started.md)  · \n  [Website](https://egm.dev)\n\nThousands of Google engineers use EGM, an open-source plugin, to draft specs with deep-context agents and stress-test them with zero-context subagents before writing code.\n\n**Elephant** holds your project context and drafts the specification. **Goldfish** reviewers examine the draft and referenced source material without the earlier conversation, helping expose unstated assumptions and blind spots. **You** make the architectural trade-offs and approve the specification before implementing.\n\nUse it when a feature needs design decisions before implementation. For a small, clear bug fix, use your normal coding workflow.\n\nChoose your coding agent. Install the complete plugin so its skills can read the reviewer prompts and document standards.\n\n```\nagy plugin install https://github.com/google/elephant-goldfish-model\nclaude plugin marketplace add google/elephant-goldfish-model\nclaude plugin install egm@egm\ncodex plugin marketplace add google/elephant-goldfish-model\ncodex plugin add egm@egm\n```\n\nStart a new session after installation. For **Claude Cowork** and **ChatGPT Work**, see the [setup guide](https://github.com/google/elephant-goldfish-model/blob/main/docs/harnesses.md).\n\nOpen your project in your coding agent and choose a feature that needs design decisions. This example adds webhook delivery: what should happen when a delivery fails, or the same event arrives twice?\n\nThe commands below are shown without a prefix. In Antigravity (agy CLI) and Claude Code, add the `/egm:` prefix, for example `/egm:egm-brainstorm`. In Codex, select the corresponding EGM skill from the skill picker. Brainstorming and review require a session that can create fresh subagents.\n\n```\n/egm-brainstorm Add webhook delivery to this project.\n```\n\nDiscuss requirements and trade-offs with the agent. Fresh reviewers challenge the proposal, and the agent saves the agreed direction in `concepts_brief.md`.\n\n```\n/egm-draft\n```\n\nThe agent turns the brief into `design_draft.md`, pausing after each section for your feedback in chat.\n\n```\n/egm-review design_draft.md\n```\n\nUse the draft path returned by the agent.\n\nFresh reviewers check the design, its clarity, its fit with the code, and its implementation plan. Choose which findings to address. When you approve the design, ask your coding agent to implement it.\n\nSee the [workflow guide](https://github.com/google/elephant-goldfish-model/blob/main/docs/getting_started.md) for the full process, including saving approved specifications under `docs/designs/`.\n\n| Role | Context | Responsibility | \n|---|---|---|\n| **Elephant** | Your project, requirements, and earlier decisions | Explore the problem and draft the specification. | \n| **Goldfish** | The document and its referenced source material, without the earlier conversation | Find missing assumptions, unclear requirements, and implementation risks. | \n| **You** | The goals and constraints that matter | Resolve trade-offs, approve the design, and decide what to build. | \n\nGoldfish reviews require fresh subagent contexts. If your agent cannot create them, EGM stops the review. A review does not replace tests or human judgment. See [supported environments and verification limits](https://github.com/google/elephant-goldfish-model/blob/main/docs/harnesses.md#support-status--contributing-other-harnesses).\n\nYou can use each skill when you need it. You do not have to follow the full sequence.\n\n| Skill | Use it to… | \n|---|---|\n|  [Brainstorm](https://github.com/google/elephant-goldfish-model/blob/main/skills/egm-brainstorm/SKILL.md) | Test an idea and write a concept brief. | \n|  [Draft](https://github.com/google/elephant-goldfish-model/blob/main/skills/egm-draft/SKILL.md) | Turn a brief or existing notes into a design document. | \n|  [Resolve](https://github.com/google/elephant-goldfish-model/blob/main/skills/egm-resolve/SKILL.md) | Work through review findings and agree on changes. | \n|  [Bootstrap](https://github.com/google/elephant-goldfish-model/blob/main/skills/egm-bootstrap/SKILL.md) | Create or update documentation for an existing codebase. | \n|  [Review](https://github.com/google/elephant-goldfish-model/blob/main/skills/egm-review/SKILL.md) | Review a design with fresh agents across five stages. | \n|  [Review loop](https://github.com/google/elephant-goldfish-model/blob/main/skills/egm-review-loop/SKILL.md) | Run a configured group of reviewers for a specific review task. | \n|  [Drift report](https://github.com/google/elephant-goldfish-model/blob/main/skills/egm-drift-report/SKILL.md) | Compare checked-in documents with the current source code. | \n\n- [Getting started](https://github.com/google/elephant-goldfish-model/blob/main/docs/getting_started.md) : the workflow and project layout.\n- [Setup](https://github.com/google/elephant-goldfish-model/blob/main/docs/harnesses.md) : installation, agent capabilities, and known limits.\n- [Architecture](https://github.com/google/elephant-goldfish-model/blob/main/docs/designs/architecture.md) : how skills and reviewers fit together.\n- [Testing](https://github.com/google/elephant-goldfish-model/blob/main/TESTING.md) : checks, execution records, and verification limits.\n- [Dave Rensin’s original essay](https://medium.com/@drensin/elephants-goldfish-and-the-new-golden-age-of-software-engineering-c33641a48874) : the ideas behind EGM.\n\nRead [contributing](https://github.com/google/elephant-goldfish-model/blob/main/CONTRIBUTING.md) for PRs, [issues](https://github.com/google/elephant-goldfish-model/issues) for bugs, and [discussions](https://github.com/google/elephant-goldfish-model/discussions) for questions.\n\n[Apache 2.0](https://github.com/google/elephant-goldfish-model/blob/main/LICENSE). This is not an officially supported Google product.", "url": "https://wpnews.pro/news/elephant-goldfish-model-skills-for-designing-and-reviewing-software-with-agents", "canonical_source": "https://github.com/google/elephant-goldfish-model", "published_at": "2026-10-30 02:09:30+00:00", "updated_at": "2026-10-30 02:46:48.909320+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "ai-products"], "entities": ["Google", "Elephant-Goldfish-Model", "EGM", "Claude Code", "Codex", "Antigravity", "Claude Cowork", "ChatGPT Work"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/elephant-goldfish-model-skills-for-designing-and-reviewing-software-with-agents", "markdown": "https://wpnews.pro/news/elephant-goldfish-model-skills-for-designing-and-reviewing-software-with-agents.md", "text": "https://wpnews.pro/news/elephant-goldfish-model-skills-for-designing-and-reviewing-software-with-agents.txt", "jsonld": "https://wpnews.pro/news/elephant-goldfish-model-skills-for-designing-and-reviewing-software-with-agents.jsonld"}}