Elephant-Goldfish Model: design documents with fresh-eyes review
google/elephant-goldfish-model
Your AI coding assistant helps you think a feature through and writes the design document with you, section by section. Then fresh reviewers who have not seen your chat read it and point out missing assumptions and risks. You decide what to change and what to build.
Pack description by its author: “Your AI coding assistant helps you think a feature through and writes the design document with you, section by section. Then fresh reviewers who have not seen your chat read it and point out missing assumptions and risks. You decide what to change and what to build.”
Software engineeringCommands you run when neededKnowledge and rules for your assistant
1. Changes your files
None of its skills edits your files.
github.com/google/elephant-goldfish-model/blob/8414530/2. Runs code on your machine
Install: Markdown skills, reviewer prompts, document standards and plugin manifests. No hooks, no scripts in the runtime, no MCP server. Reviews: Your assistant starts separate helper agents. In Claude Code they may only read files (Read, Grep, Glob).
github.com/google/elephant-goldfish-model/blob/8414530/3. Sends data
Install from github.com/google/elephant-goldfish-model or the published ChatGPT listing. The skills tell your assistant what to do: read them before installing. Give only the files a review needs, since inputs can hold confidential data.
github.com/google/elephant-goldfish-model/blob/8414530/4. Commercial use
Apache-2.0: keep the license and notices
github.com/google/elephant-goldfish-model/blob/8414530/Checked against the pack's files at commit 8414530, on Nov 13, 2026. Tap a verdict to see its source.How we checkedReport an error
The pack has changed since AgentGid checked it on Nov 13, 2026. Some details may be out of date.
Good for you if
- You want a written plan for a feature before your assistant builds it
- You have a design document, README or plan and want it checked against the real code
- You want to find where your documents and your code have drifted apart
- You use Claude Code, Codex, Antigravity or the ChatGPT or Codex desktop app
Probably not for you if
- You want a quick fix for a small, well-understood change: the authors say design and review may not pay off there
- Your assistant cannot start separate helper agents: the review stops instead of running
What its answer looks like
The concepts brief from /egm-brainstorm
# [Feature Name] * Author / Date ## Overview ## Analysis ## Goals vs. Non-Goals ## Alternatives Weighed ## Key Decisions & Staging for Next Steps
The skill's own example, from skills/egm-brainstorm/SKILL.md. More examples
What you'll notice
- Slash commands such as /egm:egm-brainstorm, /egm:egm-draft and /egm:egm-review in Claude Code; nothing starts on its own
- Many open questions first, then a concepts_brief.md file instead of a long answer in chat
- design_draft.md written one section at a time, with a pause for your approval after each
- Review reports as files, with findings marked Blocker, Major or Nit
What you need
- Claude Code, Claude Cowork, Codex, ChatGPT, Antigravity
- Free, Apache-2.0 license
- Two commands in Claude Code or Codex, or install the published listing in the ChatGPT or Codex desktop app
Words on this page
- Skill
- A set of written instructions your AI assistant follows. Installing one is like giving it a playbook.
- Skill pack
- Several skills in one GitHub repository, installed together.
- Slash command
- A command you type in the assistant's chat, starting with /.
- Plugin
- A package your assistant installs that can bring skills, commands and hooks together.
- Subagent
- A helper the assistant starts for one part of a task.
| Repository | google/elephant-goldfish-model · website |
|---|---|
| GitHub stars | 19,337 |
| Skills | 7 |
| Category | Software engineering |
| License | Apache-2.0 |
| Last updated | Nov 14, 2026 · card checked Nov 13, 2026 at commit 8414530 |
| Install counts | google/elephant-goldfish-model on skills.sh, Vercel's skills directory, which shows installs and security audits per skill |
google/elephant-goldfish-model popularity over time
What Elephant-Goldfish Model does
- Elephant: the main assistant keeps your project, requirements and earlier decisions, and drafts the document
- Goldfish: reviewers start fresh, read only the document and the files it points to, and report gaps
- You: settle trade-offs, approve the design and decide what gets built
- Plain English, no flattery: the drafting skills tell the assistant to stress-test ideas and lead with what is wrong
- Big texts go to files (brief, draft, reports), not into the chat
- When a design choice has more than one sound answer, it shows up to three options and asks you to choose
From the pack's SKILL.md files, summarised by AgentGid.
The 7 skills, explained
egm-brainstorm
Questions you about a rough idea, has fresh reviewers challenge it, and writes a short brief
You give
- The feature idea in a sentence or two
- Answers to open questions: the problem, who uses it, hard limits, edge cases, how success is measured
- A go-ahead once the summary of the idea is right
You get
- concepts_brief.md: overview, analysis, goals and non-goals, alternatives weighed, key decisions
- 2 to 3 main objections from the reviewers, named in chat
- A choice: share the brief for team feedback, start drafting, keep refining, or stop
egm-draft
Writes the design document with you, one section at a time
You give
- concepts_brief.md from brainstorm, or your own notes file
- Approval or changes after each section
- Your choice when a design question has more than one sound answer
You get
- design_draft.md, section by section: overview and goals, technical plan, alternatives, implementation
- A final document saved under docs/designs/ (or a folder you confirm)
egm-review
Has fresh reviewers check a design document in five stages
You give
- Path to the design document, README or plan
- Optional extra files for the reviewers to read
- A review mode: step by step, one combined report, the feature part only, or the implementation plan only
You get
- zoom_out_report.md, comprehension_report.md, critic_report.md, intermediate_zoom_out_report.md, implementation_readiness_report.md
- consolidated_review_report.md in the combined mode
- Findings you can accept, reject or discuss before anything is edited
egm-resolve
Sorts review findings into fix, contest, ask you, or skip, and edits the document only after your answer
You give
- The reviewed document
- The list of findings (passed by the review)
- Your decision: auto accept, continue, item by item, or your own instruction
You get
- goldfish_review_report.md with each finding, its severity and a suggested edit
- Accepted fixes applied to the document
egm-review-loop
Runs the reviewers you choose against standards you choose, round after round, for any document
You give
- The document and the part to review (or the full file)
- Which reviewers and which standards
- Files or links the reviewers should read
You get
- Review rounds until two clean rounds in a row, at most 5
- goldfish_review_report.md after each round
- A short list of open findings if it stops at the cap
egm-bootstrap
Writes or updates a README in each code folder of an existing project, describing what it offers and how to test it
You give
- The folder to document
- Your approval if it finds more than 15 folders
You get
- README.md files (or EGM.md, if you want your own READMEs left alone) with each folder's interface, rules and side effects
- A top-level README that links the folder READMEs
egm-drift-report
Compares your documents with the current code and lists only where they differ
You give
- The folder with design documents or READMEs
You get
- {date}_DRIFT_REPORT.md: one section per document, the differences or "No drift detected"
See it in action
The skill defines what its answer looks like. Its own example:
# [Feature Name] * Author / Date ## Overview ## Analysis ## Goals vs. Non-Goals ## Alternatives Weighed ## Key Decisions & Staging for Next Steps
The skill defines what its answer looks like. Its own example:
### [Issue 1] [ReviewerPersona] - Severity: [Blocker | Major | Nit] * Affected Scope: [Lines Start-End] ([Sub-header]) * Discrepancy Description: ... * Rationale: ... * AI Suggestion: [CONTEST | ESCALATE | PROPOSE FIX | DECLINE] * Justification: ... * Proposed Edit: Location / Change
The skill defines what its answer looks like. Its own example:
## <DESIGN_DOC_NAME> <Bulleted list of specific divergences found> <Or "No drift detected">
When to reach for it
| You want | Use |
|---|---|
| To test a rough feature idea before writing anything | /egm:egm-brainstorm |
| To turn a brief or notes into a design document | /egm:egm-draft |
| A fresh check of a design doc | /egm:egm-review <file> |
| Your own choice of reviewers and standards for any document | /egm:egm-review-loop |
| READMEs for every code folder of an existing project | /egm:egm-bootstrap <folder> |
| To see where documents and code no longer match | /egm:egm-drift-report <folder> |
Skills in google/elephant-goldfish-model
| Skill | In plain words | Changes files? | What it does (author) |
|---|---|---|---|
egm-bootstrap/egm:egm-bootstrap <folder> |
Writes or updates a README in each code folder of an existing project, describing what it offers and how to test it | No | Bootstrap directory README.md files across a repository subtree.npx skills add google/elephant-goldfish-model --skill egm-bootstrap |
egm-brainstorm/egm:egm-brainstorm <idea> (Claude Code, agy CLI); skill picker in Codex |
Questions you about a rough idea, has fresh reviewers challenge it, and writes a short brief | No | Interactive ideation and brainstorming skill for the Elephant & Goldfish Model (EGM). Helps turn half-baked concepts into structured hypotheses via clarifying questions, review, and iterative refinement. Produces an RFC summary for human review or more detailed design with egm-draft.npx skills add google/elephant-goldfish-model --skill egm-brainstorm |
egm-draft/egm:egm-draft [brief or notes file] |
Writes the design document with you, one section at a time | No | Drafts a design document section-by-section using the concepts brief from egm-brainstorm (or user notes). This is the second step in the Elephant-Goldfish (EGM) feature flow. Use when you are ready to write a design doc after aligning on requirements + solution.npx skills add google/elephant-goldfish-model --skill egm-draft |
egm-drift-report/egm:egm-drift-report <docs folder> |
Compares your documents with the current code and lists only where they differ | No | Runs drift detection on all design documents to check for document and code consistency. Produces an actionable report with differences. Use when the user asks to detect drift or check for consistency across all design documents or READMEs in a directory.npx skills add google/elephant-goldfish-model --skill egm-drift-report |
| egm-resolve | Sorts review findings into fix, contest, ask you, or skip, and edits the document only after your answer | No | The Elephant resolution skill. Ingests flat deficiency reports from concurrent Goldfish reviewers, applies automated fixes for PROPOSE FIX items via UI multiple-choice gates, and presents interactive negotiation/re-sync checkpoints.npx skills add google/elephant-goldfish-model --skill egm-resolve |
egm-review-loop/egm:egm-review-loop <document> |
Runs the reviewers you choose against standards you choose, round after round, for any document | No | Executes the Antagonistic-Review process from the Elephant-Goldfish Model (EGM) via an automated multi-agent review loop to audit documents against qualitative standards and structural templates, and automatically apply approved fixes. Use when reviewing design docs, READMEs, feature…npx skills add google/elephant-goldfish-model --skill egm-review-loop |
egm-review/egm:egm-review <design doc> |
Has fresh reviewers check a design document in five stages | No | Orchestrates the EGM Goldfish Review Protocol. Used as part of the EGM process to perform design document reviews. Executes five review phases: Initial Zoom Out Review, Comprehension Review, Design Critique, Intermediate Zoom Out Review, and Implementation Readiness.npx skills add google/elephant-goldfish-model --skill egm-review |
The author's descriptions come from each skill's SKILL.md; the plain-words column is AgentGid's.
Good to know before you install
- Not an officially supported Google product, as the README and the plugin listing say, even though it is published under the google organisation.
- Nothing starts by itself: every skill is turned off for automatic use, so you call it by name.
- Install the whole plugin. Copying only the skills folder breaks it, because the skills read the reviewer and standards files.
- It writes files (briefs, drafts, reports, READMEs) into your working folder; you can name another folder.
- A review does not replace tests or your own judgment.
- In Claude Code, use version 2.1.271 or later so reviewers do not see your project instructions file.
Before you start
Works in
- Claude Code: plugin
- Codex CLI: plugin
- ChatGPT and Codex desktop: plugin listing
- Antigravity CLI: plugin
- Claude Cowork: plugin
- omp, Muse: plugin
- Cursor (experimental): plugin
- GitHub Copilot CLI (experimental): plugin
You will need
- A host that can start fresh helper agents: Brainstorm and review need separate reviewers
- Claude Code 2.1.271 or later: Reviewer isolation from CLAUDE.md
- Local executor (Codex and ChatGPT): Reviews need it; cloud support is partial
Steps only you can do
- In the ChatGPT or Codex desktop app, install the published EGM listing yourself (https://chatgpt.com/plugins/plugins_6aeb90f47ec88191a3c107e01d3eb357)
Then: The agent can use the EGM skills in a new local session - Approve full-tree bootstrapping when egm-bootstrap finds more than 15 folders (skills/egm-bootstrap/SKILL.md)
Then: The agent starts parallel workers to write the READMEs - Choose which links outside the working folder or web pages the reviewers may read (skills/egm-review/SKILL.md (Step 1.3))
Then: The agent saves approved pages as files and passes them to the reviewers - Approve the design; then ask your assistant to build it (README.md)
Then: The agent implements the approved design
Mistakes to avoid
Install only the skills folder with a skills installer
Install the whole plugin for your agent
The skills read the reviewer prompts and document standards at run timeRun a review in an assistant that cannot start fresh helper agents and expect a result
Use a host that can start separate agents (Claude Code, Codex, Antigravity CLI)
The review skills stop rather than review in the same conversationIn Claude Code, call egm-resolve or egm-review-loop through the Skill tool
Start the main skills with /egm:egm-...; inside a review the assistant reads the other SKILL.md and follows it
The skills are hidden from automatic use and the Skill tool refuses themExpect /egm-drift-report to bring code and documents back in line
Read the report, then decide whether to change the document or the code
It only lists differencesUse an older Claude Code and treat reviews as isolated
Use Claude Code 2.1.271 or later
Older versions are not verified for reviewer isolation
Which one to pick
| If you need | Pick |
|---|---|
| You want a design document drafted with you and checked by fresh reviewers before anyone builds | This pack |
| You want a full coding workflow that plans, writes tests first and reviews the code it writes | obra/superpowers |
| You want approved specs turned into long autonomous implementation runs | gotalab/cc-sdd |
Works well with
- claude-code: Plugin with read-only reviewer adapters; the authors run end-to-end tests on it
- openai-codex: Codex CLI plugin; reviews need a local executor
- antigravity-cli: Installs with agy plugin install and has named reviewer subagents
- claude-cowork: Installs from the Cowork plugin panel; cloud checks still open
May I use it like this?
| Use | Answer | Condition |
|---|---|---|
| In a commercial app | Yes | Apache-2.0: keep the license and notices |
| Passing the files on | With conditions | Apache-2.0: include a copy of the license and keep notices; no trademark rights |
| Work for clients | Yes | Apache-2.0 |
What leaves your machine
- Requests for web pages linked from the reviewed document → The linked sites, Only for links you approve during a review. Opt out: Do not approve web links; the review goes on without them
- Package downloads by a project formatter → A package registry, Only if egm-bootstrap runs a markdown formatter your project already configures. Opt out: Deny the command in your assistant's permission prompt
Install Elephant-Goldfish Model
Official commands from the repository
claude plugin marketplace add google/elephant-goldfish-model claude plugin install egm@egm
Start a new session after installing. If needed, enable it with claude plugin enable egm@egm. Commands use the /egm: prefix.
codex plugin marketplace add google/elephant-goldfish-model codex plugin add egm@egm
Start a new session, then pick EGM skills from the skill picker.
Install the published EGM plugin listing in the desktop app (desktop only).
agy plugin install https://github.com/google/elephant-goldfish-model
Keep the /egm: prefix on commands.
Install from claude.ai/plugins or the Cowork plugin panel via the marketplace google/elephant-goldfish-model.
omp plugin marketplace add google/elephant-goldfish-model omp plugin install egm@egm
Skills are called as /skill:egm-<command>.
Or with the open-source skills CLI: npx skills add google/elephant-goldfish-model. Skills are instructions your agent follows: read them before installing.
What it installs and runs on your machine
Read in the code at commit 8414530, Nov 13, 2026
| When | What happens |
|---|---|
| Install | Markdown skills, reviewer prompts, document standards and plugin manifests. No hooks, no scripts in the runtime, no MCP server. |
| Reviews | Your assistant starts separate helper agents. In Claude Code they may only read files (Read, Grep, Glob). |
| egm-bootstrap | Your assistant runs local commands to list folders and may run a markdown formatter the project already has set up (for example npx prettier), which can reach a package registry. |
| Links in a reviewed document | Links that point outside the working folder or to the web are listed and need your approval; approved web pages are saved as files in the working folder for the reviewers. |
| Repository only | Development tooling makes network calls (a PR triage script in .github/scripts, the website's star counter and analytics check in site/). It is not part of the skills and never runs from them; the authors leave it out of the runtime release packages. |
Settings
EGM artifact directory: <path>: A line in your prompt, CLAUDE.md or AGENTS.md that sends briefs, drafts and reports to another folder.egmignore: Folders and files egm-bootstrap skips, like .gitignore
Quick answers
- Is it free?
- Yes. Apache 2.0 license, no EGM account or keys.
- Is it an official Google product?
- No. The README says it is not an officially supported Google product.
- Does it send my files anywhere?
- The plugin has no service, analytics or telemetry of its own. Your assistant's host handles your files under its own terms.
- Does it write code?
- No. It writes briefs, design documents, READMEs and review reports. You ask your assistant to build the approved design.
- Can I use only one skill?
- Yes. The authors say you can use each skill when you need it, without the full sequence.
- Does it work in Cursor or GitHub Copilot?
- Only as experimental adapters that the docs mark as unverified.
Checked by AgentGid on Nov 13, 2026 against commit 8414530 of the pack's own files. agent.md for AI agents · Report a mistake
More software engineering skill packs
| # | Skill pack | Skills | GitHub stars | Updated |
|---|---|---|---|---|
| 1 |
mattpocock/skills Your AI coding assistant gets a set of working habits for building software. It interviews you about a plan before it builds, tur…
|
39 | 325K | 2026-11-13 |
| 2 |
obra/superpowers Your coding assistant follows one fixed way of working. Before it builds anything, it asks what you really want and writes a shor…
|
15 | 318K | 2026-11-03 |
| 3 |
multica-ai/andrej-karpathy-skills Your AI coding assistant gets four rules: think before coding, keep it simple, change only what you asked for, and set a clear go…
|
1 | 233K | 2026-04-20 |
| 4 |
DietrichGebert/ponytail Your AI coding assistant stops overbuilding. It does what you asked in the simplest way that works, reuses what your project alre…
|
6 | 191K | 2026-10-29 |
| 5 |
Shubhamsaboo/awesome-llm-apps This repo is mostly a big collection of AI apps and tutorials. Its agent_skills folder holds seven separate skills you install on…
|
10 | 145K | 2026-11-15 |
| 6 |
addyosmani/agent-skills Your AI coding assistant works like a careful senior engineer. It writes down what to build before coding, splits the work into s…
|
25 | 118K | 2026-11-14 |