Skill

brainstorm

Run a multi-turn brainstorm between OpenAI and Gemini on a user-supplied topic, then return the transcript for synthesis. Use when the user wants to "brainstorm with other models", get cross-model perspectives, debate an idea, or explore a topic from multiple angles. Supports a research mode where both models use web search.

npx claudepluginhub moebit/claude --plugin multi-llm

Invocation

How this skill is triggered — by the user, by Claude, or both

Slash command

/multi-llm:brainstorm <topic>

User invocable

Model invocable

Inline context

Default effort

Argument hint<topic>

Tool Access

This skill is limited to the following tools:

Bash(uv run *)

Context Preview

The summary Claude sees in its skill listing — used to decide when to auto-load this skill

This skill runs a back-and-forth dialogue between OpenAI and Gemini on `$ARGUMENTS`, prints the transcript to stdout, and lets you (Claude) synthesize the result for the user.

Supporting Files

scripts/brainstorm.py

SKILL.md

56 lines · ~653 tokens

Similar Skills

cache-components

139.5k

Guides Next.js Cache Components and Partial Prerendering (PPR): 'use cache' directives, cacheLife(), cacheTag(), revalidateTag() for caching, invalidation, static/dynamic optimization. Auto-activates on cacheComponents: true.

cache-components

pdf

131.6k

Processes PDFs: extracts text/tables/images, merges/splits/rotates pages, adds watermarks, creates/fills forms, encrypts/decrypts, OCRs scans. Activates on PDF mentions or output requests.

11 files

document-skills

Stats

LanguageHTML

Stars0

MaintenanceExcellent

Last CommitMay 13, 2026

Actions

View Source View Plugin View on GitHub View README

Help us improve

Share bugs, ideas, or general feedback.

Stats

Actions

Help us improve

Share bugs, ideas, or general feedback.

Brainstorm with OpenAI and Gemini

This skill runs a back-and-forth dialogue between OpenAI and Gemini on $ARGUMENTS, prints the transcript to stdout, and lets you (Claude) synthesize the result for the user.

How to invoke

Default (3 rounds, no web search):

uv run "${CLAUDE_SKILL_DIR}/scripts/brainstorm.py" --topic "$ARGUMENTS"

Research mode — both models use web search:

uv run "${CLAUDE_SKILL_DIR}/scripts/brainstorm.py" --topic "$ARGUMENTS" --research

Pro tier — when the user asks for the "Pro models", "best models", "highest quality", "deep thinking", or similar, add --pro to use gpt-5.5-pro and gemini-3.1-pro-preview instead of the cheaper defaults:

uv run "${CLAUDE_SKILL_DIR}/scripts/brainstorm.py" --topic "$ARGUMENTS" --pro

Other flags:

--turns N — number of rounds (default 3; each round is one OpenAI turn + one Gemini turn)
--start gemini — Gemini opens (default: OpenAI opens)
--openai-model NAME — override OpenAI model (default gpt-5.5, gpt-5.5-pro with --pro, or $BRAINSTORM_OPENAI_MODEL)
--gemini-model NAME — override Gemini model (default gemini-3.1-flash-lite-preview, gemini-3.1-pro-preview with --pro, or $BRAINSTORM_GEMINI_MODEL)

If the user expressed intent like "research it" or "look it up", add --research. If they asked for "a quick exchange" or "just one round", set --turns 1. If they asked for "the best" or "Pro tier", add --pro. These flags compose freely.

Required environment variables

OPENAI_API_KEY — OpenAI API key
GEMINI_API_KEY (or GOOGLE_API_KEY) — Google AI API key

If either is missing, the script exits non-zero with a clear message. Relay it to the user and ask them to export the missing key before re-running.

After the script runs

Don't just paste the transcript back. Synthesize:

Rank the strongest ideas across both models.
Note where OpenAI and Gemini disagreed or emphasized different angles.
Call out any novel point the user probably hadn't considered.
End with a concrete next step or a follow-up question.

The raw transcript is already on the user's screen — your value is judgment on top of it.