Curate

Find and compare relevant options using current evidence. Scale research depth to the decision's stakes, uncertainty and cost. General research that does not involve an options comparison does not require this skill. Keep reusable evidence with its relevant plan or designated deliverable; use .work/temp/<task>/ for scratch research under the workspace rules.

Compare options

1. Identify the job, audience, constraints and decision criteria from the request and available context. Include named candidates; ask a focused question when an unresolved requirement would materially change the options or decision, and continue research that does not depend on the answer.
2. Choose the route. For discovery, search sources suited to the decision for a candidate pool before narrowing it. Look beyond familiar options when credible alternatives may fit. For a comparison of known options, start with those options and the evidence relevant to the stated criteria. Broaden only if doing so could materially improve the decision and the requested scope allows it.

Check the evidence

3. For each serious candidate, inspect evidence that substantiates its fit and constraints. Use internal project evidence for internal options and primary sources such as official documentation, repositories, release notes or vendor information for external claims. Verify time-sensitive details such as compatibility, availability, pricing, licensing and maintenance when they matter to the choice.
4. Investigate adoption, engagement or stability when they could change the recommendation. Choose signals appropriate to the decision, such as maintenance history, usage reports, or independent customer evidence. Record what each signal measures and its date or period; popularity alone does not establish quality or stability.
5. Compare candidates using the same relevant criteria. Prioritize fit, material risks and costs. Distinguish current observations from older information and public claims from independently supported facts. Mark missing evidence as unknown; never invent metrics or imply independent verification where none exists.

Present the options

6. Present the relevant candidates, why each fits, its key tradeoff, and the evidence supporting the assessment. Link sources near the claims they support and date time-sensitive evidence. Distinguish observation from judgment.
7. Give a recommendation when evidence and requirements support one, explain when another option would be better, and leave the decision with the human. When a material requirement is unresolved, offer conditional recommendations and ask about the decision that would change the choice. If current sources or important evidence are unavailable, state the limitation and what it prevents you from determining; offer a clearly provisional comparison when useful, without presenting it as current research.

Done: The human can compare relevant options using current, attributed evidence about the criteria that matter, with uncertainty and tradeoffs made clear.