Guide · Agent skill
How to use the EGM Brainstorm agent skill (egm-brainstorm)
Updated · written with AI from the project's own documentation and checked against our catalogue data
egm-brainstorm is an agent skill for the Elephant & Goldfish Model (EGM) project. It turns a rough feature idea into a written concept brief after a multi-turn interview and a review from several reviewer perspectives. This guide explains what a skill is, how to install this one, how to trigger it, and what it needs before you start.
EGM Brainstorm Ideationgoogle/elephant-goldfish-model · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Get the egm-brainstorm ZIP from the directory page.
- 2Unzip into a skills folder. Place the contents in ~/.claude/skills/ for personal use, or in .claude/skills/ inside a project.
- 3Check your harness. Confirm you are using a harness with a subagent tool, such as Claude Code, Codex, or Antigravity.
- 4Ask for an EGM brainstorm. Name the skill and describe your rough feature idea in plain language.
- 5Answer the clarifying questions. Respond to the open-ended questions over several turns until the problem and boundaries are clear.
- 6Confirm readiness. Review the working hypothesis and confirm it so the Goldfish reviews can start.
- 7Open concepts_brief.md. Find the file the agent wrote, share it with peers, or pass it to /egm-draft.
What is an agent skill and what does this one make the agent do?
An agent skill is a folder of instructions that an AI coding agent loads when you need it. This skill sets the agent up as a structured brainstorming partner for planning software features in the EGM project.
The agent first asks you open-ended questions across several chat turns about the problem, its constraints, edge cases, how the problem is handled today, and how success would be measured. It does not rush into review. Once the problem and boundaries are clear and you confirm readiness, it gets critiques from several reviewer lenses, including contrarian/YAGNI, cost and token economics, UX and developer ergonomics, migration and backward compatibility, observability, and long-term maintenance.
- Output is a file named concepts_brief.md in the working directory (or your harness's artifact directory), not a chat essay.
- The agent tells you in chat where the file is.
- The brief can be shared with peers or passed to the companion /egm-draft skill for a detailed design.
How do you install the egm-brainstorm skill?
Download the egm-brainstorm ZIP from the directory page. Unzip it into one of two locations: your personal skills folder, or a project-specific folder if you want the skill limited to one repository.
Alternatively, in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal: ~/.claude/skills/
- Project: .claude/skills/ inside your project folder
~/.claude/skills/How do you trigger the skill with a natural request?
The skill's frontmatter sets disable-model-invocation to true, so the agent is not meant to pick it up on its own. Name the skill or clearly ask for an EGM brainstorm when you want it. The phrasings below are examples you can adapt, not required wording.
Start with a concept that may still be vague. You do not need a finished spec. The agent will ask questions before it creates anything.
- Use the egm-brainstorm skill to explore a retry queue for our EGM ingestion pipeline.
- Brainstorm a feature idea with me using egm-brainstorm. I have a rough concept and want it pressure-tested.
- Run egm-brainstorm on this idea and produce a concept brief I can share with the team.
What does the skill need before it runs?
The skill depends on a harness that supports subagents. The reviews must run in a fresh context, not in the same conversation. The skill names these tools: Claude Code Agent tool, Codex spawn_agent, or Antigravity invoke_subagent.
The skill also references companion files: agents/<persona>.md persona files and docs/claude.md, docs/codex.md, and docs/harnesses.md. The /egm-draft skill is optional and is only needed for the detailed design step.
- A harness with subagent support (Claude Code, Codex, or Antigravity as named above)
- Companion files: agents/<persona>.md, docs/claude.md, docs/codex.md, docs/harnesses.md
- Optional: the /egm-draft companion skill
- If the agent has no subagent tool, the skill says it will stop rather than run a same-context review
What safety notes should you know about?
According to the directory listing, this skill runs no scripts and makes no network calls. Its output is a Markdown file written to your project or artifact directory.
Because the skill writes files, check the working directory before you start so the brief lands where you expect.
- Runs scripts: no
- Uses network: no
- Writes its output as a file in the working directory or harness artifact directory
What tips help you get a useful brief?
Bring the real problem, not a proposed solution. The agent's questions focus on the problem statement and boundaries, so a clear pain point gives it more to work with.
Answer the questions with specifics such as current workarounds, constraints, and what success looks like. Expect the critiques to lead with weaknesses; the skill is designed to stress-test rather than agree.
- Name constraints early, such as offline operation or a specific platform or dialect.
- Say how the problem is handled today and why that fails.
- Read the readiness summary carefully before you confirm, because the reviews only start after you do.
- Take the finished concepts_brief.md into /egm-draft when you need a fuller design.
What people use it for
- Pressure-testing a feature idea for the EGM project before writing a design doc.
- Producing a written RFC-style summary from an early, vague concept for team review.
- Checking a proposed feature against cost, migration, and maintenance concerns before committing to it.
- Turning a rough idea into a brief that the /egm-draft companion skill can expand into a detailed design.
Frequently asked questions
Is the egm-brainstorm skill free to use?
The directory lists the skill under the Apache-2.0 licence. Check the licence file in the download for the terms that apply to your use.
How do I install egm-brainstorm in Claude Code?
Download the ZIP, unzip it into ~/.claude/skills/ for personal use or .claude/skills/ inside your project. You can also upload it in the Claude apps under Settings, then Capabilities, then Skills.
What does egm-brainstorm need to run?
It needs a harness with a subagent tool, such as Claude Code Agent tool, Codex spawn_agent, or Antigravity invoke_subagent. It also needs its companion persona and docs files, and /egm-draft is optional.
Does egm-brainstorm run scripts or use the network?
According to the directory listing, it runs no scripts and uses no network. It writes its output as a file.
Why won't the agent give me a chat summary?
The skill requires the brief to be written to a file and forbids presenting large briefs in the chat pane. The agent will tell you where the file is.
What should I use instead if I just want a quick chat brainstorm?
Use a general chat session without this skill. The directory notes that egm-brainstorm is not suited to quick brainstorms with no files, or to projects unrelated to EGM.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use egm-brainstorm if you work on the EGM project and want a feature idea pressure-tested and written up as a shareable brief. Pick something else if you want a quick chat brainstorm with no files, work outside EGM, or use a harness without subagent support.
Download the skill