Research brief

Write an assignment another researcher can execute without inheriting the conversation. Match its depth to the question rather than forcing a paragraph, fixed subquestion count, or exhaustive research process.

Establish the job

Identify the question and end use: a decision, explanation, landscape, verification, or learning objective. Use known context and constraints. Ask when a missing end use or boundary would materially change the research; otherwise state reasonable assumptions and proceed.

Include only context the researcher needs. Specify relevant scope, exclusions, timeframe, geography, population, resource limits, and audience. Do not invent deadlines, budgets, or acceptance thresholds. Mark unresolved preferences explicitly when the brief can still be useful.

Define evidence and comparisons

Break the question into the subquestions needed to answer it. Choose source standards appropriate to each claim: primary documentation for capabilities, applicable records for historical facts, and actual participant accounts for experience. Treat vendor claims as claims until supported; source prestige does not compensate for the wrong population or time period.

For comparisons, define the common criteria, units, conditions, and relevant alternatives. Require material differences in coverage to be visible rather than forcing a ranking from incomparable evidence. Include contrary evidence and realistic alternatives to a favored hypothesis.

Require source pointers for consequential claims and distinguish facts, inferences, assumptions, and unresolved questions. Ask the researcher to trace repeated claims to their originating evidence where independence matters. Specify how to handle disagreement: compare definitions, dates, methods, and directness rather than resolving it by majority vote.

Make completion observable

Define sufficient coverage for the actual question. For a decision brief, each decisive criterion should have supported evidence or an explicit gap and its effect on the decision. For an exploratory landscape, state the search scope and coverage limits; do not promise exhaustive discovery without a bounded universe.

Require a final gap check focused on claims that could change the answer. Stop when the agreed scope and evidence requirements are met, further searching is unlikely to materially change the result, or the stated resource limit is reached. A time limit does not turn an unresolved claim into a finding. Negative searches should describe where and how the researcher looked, not imply that no evidence exists anywhere.

Deliver a runnable assignment

Return the brief in a format suited to its complexity, with the expected deliverable and what the final answer must make clear. Keep supported recommendations distinct from factual findings and request usable source links or locations. Do not perform the full investigation, dispatch another agent, or contact a researcher unless execution is also requested. If research is already authorized, use the brief as working context and continue rather than adding an approval gate.