Aglet

Prioritize Evidence Collection for an Agent Task

Prioritizing missing evidence means choosing which observation will most improve a human decision. Rank gaps by the consequence of acting without them, uncertainty in the claim, freshness, and collection effort. Keep low-impact context separate from evidence that changes whether the result can be accepted.

Decide where the work belongs

  1. Rank unsupported claims

    For each claim, record the decision it influences, the confidence without evidence, the possible outcomes, and the cost of being wrong. Give priority to evidence that tests a material boundary or distinguishes competing explanations. Do not rank a gap by artifact count alone.

  2. Assess evidence yield

    Describe the source, time window, comparison, or fixture that could address each gap, and whether it is available and trustworthy. Prefer a bounded artifact that tests several related claims when its provenance is clear. Mark freshness or access uncertainty beside the item.

  3. Choose the queue action

    Select collect a named artifact, narrow the claim, proceed with an explicit limitation, request human context, or stop review. Include owner, format, cutoff, and a trigger for reprioritization. Keep unsupported material claims out of the accepted output until their evidence standard is met.

What to carry forward

Return a ranked evidence queue with claim, decision consequence, missing observation, source, owner, freshness, and acceptance trigger. The top item should explain what review outcome it can change. Mark the recommendation provisional when the highest-impact claim remains untested.

Keep the decision with the work.

Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.

Create an account See the product workflow