# Research Planning

> Scope a research question, decide what evidence would answer it, and plan the search before diving in. Use when starting any research task, so effort goes to the questions that matter instead of wandering.

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

---


# Research planning

Research that starts with "let me look into this" wanders, collects
tangents, and stops when tired rather than when answered. Planning first
(what exactly am I answering, what would answer it, where would I find it)
turns research from a random walk into a directed search.

## Method

1. **Sharpen the question until it is answerable.** "Is X good?" is
   unanswerable; "Does X reduce Y for users like Z, at what cost?" can be
   researched. State precisely what you need to know and, crucially, what
   decision the answer informs: the decision bounds the depth and tells
   you when you have enough (see hypothesis-driven-work).
2. **Define what evidence would settle it.** Before searching, decide what
   a convincing answer looks like: what facts, data, or sources would move
   you, and what would change your mind. This prevents motivated searching
   (collecting only what confirms a hunch) and tells you what to look for
   (see the falsifiability ethic in hypothesis-driven-work).
3. **Decompose into sub-questions.** Break the main question into the
   pieces you can actually look up, ordered so early answers shape later
   searches. A complex question is a tree of answerable ones; research the
   tree, not the vague whole (see agent-task-breakdown for the analogous
   decomposition).
4. **Plan the sources by type.** Match each sub-question to where its
   answer lives: primary sources (papers, docs, data) for facts,
   practitioners for experience, multiple independent sources for contested
   claims (see source-evaluation, literature-review). Know which questions
   need authoritative sources versus a quick lookup.
5. **Set a depth budget and a stopping rule.** Decide up front how deep
   this warrants (a quick check vs a deep dive) and when to stop:
   typically when new sources stop changing the answer (saturation) or the
   decision is adequately supported. Research expands to fill available
   time without a stopping rule (see the time-box in root-cause-analysis).
6. **Stay open to reframing.** If early findings show the question was
   wrong or the real issue is elsewhere, update the plan rather than
   grinding through the original one. Good research often changes its own
   question; hold the plan firmly but not rigidly.

## Boundaries

- Planning is proportional to stakes: a quick fact needs no plan; a
  decision-shaping investigation does. Do not over-plan a five-minute
  lookup.
- A plan is a starting hypothesis about how to find the answer, not a
  script; the search will reveal better paths (see deep-research's
  iterative loop).
- Planning what to research does not evaluate what you find; that is
  source-evaluation and fact-checking, applied to every result.