Self-contained research brief

Draft one self-contained research brief paragraph that another human or AI researcher can execute without a follow-up exchange.

Collect inputs

Set context

Frame one question

Add sub-questions

Rule 01

The output is the assignment, not the answer

Rule 02

One paragraph must be enough context for a cold researcher

Rule 03

Primary sources outrank forums and social posts

Rule 04

Contradictions and single-source claims require a gap round

Execution Map

Core lanes from input to output, aligned with the command's real execution path.

Start

Input, scope, route

1 Collect inputs Extract the decision, audience, deadline, known facts, constraints, intended use, and relevant project context.
2 Set context Open the paragraph with plain-English project context and situation for a reader with no prior knowledge.

Work

Agent / skill execution

3 Frame one question Define one research question and name the decision that the answer will inform.
4 Add sub-questions Include three to six numbered inline sub-questions that cover the decision without mixing unrelated missions.

Verify

Gate, review, validation

5 Specify evidence rules State include/avoid constraints, source hierarchy, contradiction handling, gap round, and completion bar.
6 Define finding format Require each finding to carry a URL, exact claim, and one-line relevance to the decision, delivered as one detailed Markdown result.

Report, handoff, artifact

7 Return one paragraph Remove prefaces, headings, marketing copy, speculative solutions, and any private or fabricated facts.

Syntax, arguments, subcommands

How to invoke the skill: syntax, positional arguments, shared options, then each subcommand with its own syntax and outcome.

Syntax

/ak:research-prompt <research topic | decision>

Arguments

<research topic | decision> Required

Research topic or decision

Free-form context for one focused research assignment: project situation, decision, audience, deadline, known facts, constraints, intended use, source preferences, and non-goals. The Skill writes the brief; it does not run the research or choose an answer.

/ak:research-prompt "Decision: choose an authorization model for our multi-tenant API. Audience: architecture review. Deadline: Friday. Compare database RLS and service-layer checks; require current primary sources, security failure modes, migration cost, and unresolved gaps."

Sample Prompt

Concrete invocations for each flag, subcommand, mode, or route available in the live workflow.

Decision research brief Recommended

/ak:research-prompt whether to adopt Turbopack for this Next.js app

Use when:: Use when you need a research assignment, not the research itself.
Expected:: One self-contained paragraph that states project context, one research question, three to six inline sub-questions, source hierarchy, contradiction handling, gap round, completion bar, and per-finding URL/claim/decision-relevance requirements.

Deep-research handoff

/ak:research-prompt compare queue options for our image rendering pipeline

Use when:: Use when a human or AI researcher needs enough context to proceed without asking follow-up questions.
Expected:: A single-paragraph brief that frames the pipeline decision, separates include and avoid constraints, requires primary sources over weak signals, and asks for one detailed Markdown result with evidence-backed findings.

Source-backed product choice

/ak:research-prompt evaluate whether Polar or Stripe is a better fit for our subscription launch

Use when:: Use when the next step is to hand off a decision-focused research brief to a research runner such as ak-research.
Expected:: Exactly one paragraph that gives a cold researcher the decision, audience, constraints, source expectations, conflict rules, and output format needed to produce decision-ready findings.

Handled Scope

Research brief writing
Decision framing
Evidence hierarchy
Contradiction handling
Per-finding citation format