Exécutez n'importe quel Skill dans Manus
en un clic

Exécutez n'importe quel Skill dans Manus en un clic

Commencer

research-brainstorm

Étoiles1

Forks0

Mis à jour3 mars 2026 à 03:33

Brainstorm follow-up research ideas from papers. Use when asked to generate research directions or ideas.

Installation

Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.

Exécuter dans Manus

Source

jusjinuk

jusjinuk/internbot

Ouvrir le dépôt GitHub Voir les dépôts du créateur

Téléchargement

Exécuter dans Manus

Métiers associésSOC

Basé sur la classification professionnelle SOC

Enseignants postsecondaires en anthropologie et archéologieEnseignement et bibliothèques·SOC 25-1061

SKILL.md

readonly

name	research-brainstorm
description	Brainstorm follow-up research ideas from papers. Use when asked to generate research directions or ideas.

Research Brainstorming

When to use

User says "what could we do next based on this paper?"
User asks for research directions, follow-ups, or extensions
User wants to identify gaps in existing work

Workflow

Understand the paper deeply before brainstorming
- What's the core contribution? What's actually new?
- What assumptions does it rely on?
- What did they NOT do, and why?
- What are the stated limitations?
Identify gaps from multiple angles:
- Methodology: can the approach be applied to different domains/modalities?
- Scale: does it work at 10x/100x the tested scale?
- Assumptions: what happens when key assumptions are relaxed?
- Combination: can this be combined with orthogonal recent work?
- Failure modes: where does the method break down?
Generate 3-5 concrete research directions, each with:
- One-line description: specific and actionable
- Gap: what's missing in the current work that this addresses
- Approach sketch: 2-3 sentences on how you'd actually do it
- Feasibility: easy (weeks, existing tools) / moderate (months, some new infrastructure) / hard (6+ months, fundamental challenges)
- Resources needed: data, compute, specific expertise
Quality bar: every idea must be specific enough to start working on
- Bad: "improve the model's performance"
- Bad: "apply to other domains"
- Good: "replace the cross-attention in the fusion module with a mixture-of-experts gate, since the current design scales quadratically with the number of modalities — test on the AudioSet + ImageNet joint benchmark"
Prioritize by the intersection of novelty and feasibility — the best ideas are those that are clearly doable and clearly haven't been done

Formatting for Slack

Number the ideas
Bold the one-line description
Keep feasibility assessments honest — don't oversell
If an idea requires specific expertise the lab may not have, say so

Plus depuis ce dépôt

même dépôt

manage-schedules

jusjinuk/internbot

View, add, update, or cancel scheduled tasks. Use when asked about schedules, cron jobs, reminders, or recurring tasks.

2026-03-041

onboard

jusjinuk/internbot

Set up internbot for development. Use when setting up the project for the first time.

2026-03-041

daily-papers

jusjinuk/internbot

Review today's HuggingFace daily papers, pick the most relevant one, write a LaTeX summary report with follow-up research directions. Use when asked to check daily papers, do a paper review, or triggered by the scheduler.

2026-03-041

latex-report

jusjinuk/internbot

Write LaTeX research reports. Use when asked to write, compile, or publish a report and save it to the dedicated reports repository.

2026-03-041

add-channel

jusjinuk/internbot

Add a Slack channel to internbot's whitelist. Use when the user wants the bot to listen to a new channel.

2026-03-041

arxiv-search

jusjinuk/internbot

Search and summarize arxiv papers. Use when asked about papers, research topics, or specific arxiv IDs.

2026-03-031

name	research-brainstorm
description	Brainstorm follow-up research ideas from papers. Use when asked to generate research directions or ideas.

Research Brainstorming

When to use

User says "what could we do next based on this paper?"
User asks for research directions, follow-ups, or extensions
User wants to identify gaps in existing work

Workflow

Understand the paper deeply before brainstorming
- What's the core contribution? What's actually new?
- What assumptions does it rely on?
- What did they NOT do, and why?
- What are the stated limitations?
Identify gaps from multiple angles:
- Methodology: can the approach be applied to different domains/modalities?
- Scale: does it work at 10x/100x the tested scale?
- Assumptions: what happens when key assumptions are relaxed?
- Combination: can this be combined with orthogonal recent work?
- Failure modes: where does the method break down?
Generate 3-5 concrete research directions, each with:
- One-line description: specific and actionable
- Gap: what's missing in the current work that this addresses
- Approach sketch: 2-3 sentences on how you'd actually do it
- Feasibility: easy (weeks, existing tools) / moderate (months, some new infrastructure) / hard (6+ months, fundamental challenges)
- Resources needed: data, compute, specific expertise
Quality bar: every idea must be specific enough to start working on
- Bad: "improve the model's performance"
- Bad: "apply to other domains"
- Good: "replace the cross-attention in the fusion module with a mixture-of-experts gate, since the current design scales quadratically with the number of modalities — test on the AudioSet + ImageNet joint benchmark"
Prioritize by the intersection of novelty and feasibility — the best ideas are those that are clearly doable and clearly haven't been done

Formatting for Slack

Number the ideas
Bold the one-line description
Keep feasibility assessments honest — don't oversell
If an idea requires specific expertise the lab may not have, say so