# Research Taste

> Apply the user's research taste, priors, and critique style to research ideation, paper analysis, experiment design, evaluation planning, strategy, and technical research judgment. Use when the task is about choosing research directions, judging research ideas, designing studies, interpreting results, or deciding what work is worth doing.

- Skill: `leo1oel/research-taste` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add leo1oel/research-taste`
- Raw SKILL.md: https://api.skillmd.com/api/skills/leo1oel/research-taste/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: leo1oel (https://skillmd.com/u/leo1oel)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/leo1oel/research-taste

---


# Research taste

Apply the user's research taste when it would improve the answer: choosing directions, judging ideas, designing studies, interpreting results, or deciding what is worth doing.
Treat everything here and in the modules as guidance, not ground truth. Distinguish preferences from evidence, and push back when the evidence disagrees.

## Default stance

Prefer research that clarifies mechanisms, exposes bottlenecks, creates useful abstractions, or rules out plausible wrong directions.
Be especially careful about chasing trends without understanding why they matter, confusing clean narratives with true mechanisms, overvaluing novelty and undervaluing correctness, scaling before debugging, ignoring failed experiments, reading summaries instead of papers, producing confident prose that hides uncertainty, and making research plans that are too expensive to falsify quickly.

## Pick the module that fits the task

The detail lives in focused modules under `references/tastes/`. Read the one that fits; for open-ended research-direction work, read both.

- **Choosing what to work on** — idea generation, problem selection, predictions as taste, critiquing ideas, depth-vs-breadth strategy. Read `references/tastes/problem-selection.md`.
- **Doing the work rigorously** — experiment design, evaluation, engineering, and the artifacts to leave behind. Read `references/tastes/method-and-rigor.md`.