# Researchstack Idea Refine

> Guided repair skill for research ideas that are important but not yet submission-ready. Use to strengthen a thesis, method, mechanism, predictions, evidence path, and scope while preserving a user-chosen topic whenever possible.

- Skill: `lqf0624/researchstack-idea-refine-2` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add lqf0624/researchstack-idea-refine-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/lqf0624/researchstack-idea-refine-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: lqf0624 (https://skillmd.com/u/lqf0624)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/lqf0624/researchstack-idea-refine-2

---


<!-- AUTO-GENERATED for codex from researchstack v0.3.0. Edit source SKILL.md files, then regenerate from the source checkout. -->

# Idea Refine

Read [../researchstack/references/workflow.md](../researchstack/references/workflow.md), [../researchstack/references/conference-story-patterns.md](../researchstack/references/conference-story-patterns.md), [../researchstack/references/novelty-boundary.md](../researchstack/references/novelty-boundary.md), and [../researchstack/references/evidence-contracts.md](../researchstack/references/evidence-contracts.md).

Do not equate refinement with narrowing the topic. Diagnose what is weak first:

- problem significance,
- method delta,
- mechanism explanation,
- falsifiability,
- evidence feasibility,
- paper scope,
- venue story.

## Repair Flow

1. Preserve the fixed topic and semantic constraints unless the user authorizes a pivot.
2. Separate observation, research question, hypothesis, mechanism, prediction, rival explanation, and evidence.
3. Identify the closest method competitor and where the current method collapses or survives.
4. If the method is weak, invoke the `researchstack-method-synthesis` workflow rather than only making the claim smaller.
5. Turn surviving candidates into two to four falsifiable hypotheses with boundary conditions.
6. Define a decisive experiment where the proposed mechanism and a named rival predict different outcomes.
7. Separate the core contribution, enabling implementation, and future program branches.
8. Apply the one-paper envelope only after the method and evidence path are concrete.

Useful repairs include changing the information signal, action granularity, objective, coordination boundary, temporal model, layer interface, or failure model; adapting an established mechanism only when the target constraints require a non-obvious change; and recasting a weak system claim as a truthful measurement or artifact claim when that is the real contribution.

End with:

- revised thesis and title candidates,
- topic state and method state,
- method delta and borrowed-technique boundary,
- mechanism, prediction, rival, and falsifier,
- abstract skeleton,
- closest-work matrix,
- first decisive figure or experiment,
- must-have evidence,
- stop-doing and future-program lists.