A long agent run is easy to start and hard to read afterwards. These are the passes I make, in order.
01#Pass one: the ending
Read the last twenty steps first. Did the agent stop because it finished, because it gave up, or because it believed it finished? The third case is the most common and the most useful.
02#Pass two: the task list
If the harness keeps a checklist, diff it over time. Items that were ticked and later unticked are where the agent found out it was wrong.
03#Pass three: the first wrong turn
Binary-search the trace for the first step whose output you would reject. Everything after it is usually a consequence.
04#Pass four: what it never looked at
List the files and docs the agent could have read and didn’t. Missing context explains more failures than bad reasoning.
Placeholder post for the blog layout.