Build a useful investigation brief
Collect competing readings
Write two or more plausible interpretations of the request and the different work each would cause. Mark which words, omitted boundaries, or conflicting instructions create the split. Ask what evidence would distinguish the readings, then preserve the original phrasing without silently correcting it.
Trace setup inputs
Inspect task context, source references, time windows, expected format, constraints, and review owner. Compare what the agent would receive with what each interpretation requires. Record missing artifacts, stale references, and undefined terms at the point they affect execution.
Test a bounded rewrite
Draft a brief with one decision, scope, evidence list, constraints, output schema, and stop rule. Apply it to a small representative example and note any remaining fork in behavior. Keep unresolved authority or policy choices visible rather than resolving them by wording alone.
What to carry forward
Produce an evidence brief showing the original ambiguity, competing readings, missing setup inputs, bounded rewrite, and remaining decision points. Recommend the smallest clarification or evidence request that makes work executable. Do not claim the brief is clear merely because one reader can infer an intent.
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