Turn unanswered questions into a research backlog

A research backlog should explain why a question matters and what evidence would resolve it. Avoid a long list of vague topics without decision relevance.

Define the evidence boundary

Give the research teammate a bounded set of documents or an explicit source scope. Request references next to claims and keep observation separate from interpretation. An AI summary is not independent verification of a source. Where live web access is unavailable, work from the supplied snapshot and label its date rather than describing the findings as current.

Inputs for this workflow: Open questions from planning notes and current decisions.

Work through the task

1. Group duplicate questions.
2. Define a concrete evidence need for each.

A brief you can adapt

Write research-backlog.md with question, decision affected, existing evidence, next source and owner placeholder. Do not assign people unless named.

The filenames above are examples. Replace them with your actual inputs and destination before submitting the task.

Review the deliverable

Remove items whose answers would not affect the current scope and split questions that require incompatible methods.

Where this approach can fail

A suggested next source is a lead, not evidence already collected.

For the related desktop setup, see write task brief.