Aglet

Investigate Why an Agent Repeats the Same Work

Investigation should show why an agent or reviewer re-enters a completed path. Compare task context, retrieved artifacts, prior conclusions, and handoff state across attempts, then reproduce a small repeat. Keep missing context, moving data, and absent closure criteria as separate hypotheses.

Build a useful investigation brief

  1. Fingerprint each attempt

    For every run, record task wording, source set, time window, revision, queries, fixtures, findings, output, and open questions. Create a compact fingerprint from those fields and compare overlap. Mark which artifacts were actually available at execution time rather than assuming prior work was visible.

  2. Trace handoff visibility

    Follow the prior result into the next task setup, including summary, evidence index, unresolved items, owner, and stopping condition. Check whether a changed source or question invalidated the prior conclusion. Record exactly what was absent or different when repetition began.

  3. Reproduce one repeat

    Use a bounded prior brief and the next task context to replay the setup without adding new evidence. Compare selected sources, first question, path taken, and output. Then add only the missing handoff field or changed artifact to see whether behavior diverges.

What to carry forward

Produce a repeat-investigation brief with attempt fingerprints, handoff trace, changed context, reproduction, and tested explanations. Recommend a concrete handoff or stop-rule correction. If the problem is genuinely moving data, keep rechecks explicit and bounded instead of labeling them duplicate work.

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