research-plan

name

research-plan

description

Discuss a research problem with the user, clarify the scope, and persist the outcome as a durable research plan plus a planning history log. Use when the user wants to shape an idea, refine a research question, compare directions, define milestones, or produce a reusable plan for implementation and optimization. Triggers on: research plan, refine this idea, discuss a paper idea, build a research plan, turn this into a project plan, define the research scope.

Research Planning

Plan the research direction for: $ARGUMENTS

Read ${CLAUDE_SKILL_DIR}/references/planning-artifacts.md before updating the planning files.

Start every invocation by reading:

1. runtime/RESEARCH_STATE.json if it exists
2. research/plan.md if it exists
3. research/plan-history.md if it exists

Purpose

Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one-by-one. For each question, provide your recommended answer.

Ask the questions one at a time.

If a question can be answered by exploring the codebase, explore the codebase instead. This skill is not a one-off brainstorm. It creates the smallest planning artifact set that the next session can read directly without repeating the same discussion.

Required Artifacts

Maintain only these files under research/:

• research/plan.md — current plan and handoff document
• research/plan-history.md — append-only history of planning revisions

Workflow

1. Read any existing planning artifacts first.
2. Compare the current user request against the existing plan:
   • If aligned, refine the existing plan.
   • If materially different, update the plan and record the scope change in plan-history.md.
3. Ask only the minimum clarifying questions needed to remove dangerous ambiguity.
4. Write or update:
   • the objective,
   • the problem definition,
   • the thesis,
   • milestones,
   • baseline definition,
   • evaluation plan,
   • risks,
   • non-goals,
   • implementation handoff,
   • optimization handoff.
5. Append a planning entry to research/plan-history.md.
6. Update runtime/RESEARCH_STATE.json if it exists with a short summary of the current goal, status, next step, and key files.

Output Contract

research/plan.md

Must contain:

• objective,
• problem definition,
• thesis,
• milestones,
• baseline definition,
• evaluation plan,
• risks,
• non-goals,
• implementation handoff,
• optimization handoff.

research/plan-history.md

Each append should include:

• timestamp,
• what changed,
• why it changed,
• previous assumptions,
• new assumptions,
• next action.

Planning Rules

• Prefer a compact, testable thesis over a broad wishlist.
• Make hypotheses falsifiable.
• Include default metrics if the user did not specify any.
• Treat literature validation as a follow-up task unless it is the immediate focus.
• The plan should be ready for /research-implement and /research-optimize without re-planning from scratch.
• Use Implementation Handoff and Optimization Handoff to make downstream consumption explicit.
• Keep runtime/RESEARCH_STATE.json lightweight. It should orient the next agent quickly, not duplicate the full plan.

Boundaries

• Do not start coding in this skill.
• Do not launch experiments in this skill.
• Do not leave planning results only in the conversation; persist them to files.