Collection
Make the agent loop finish, recover and stay inspectable
Turn current 2025-2026 agent engineering into practical controls for bounded loops, context, memory, checkpointing, evaluation and trajectory monitoring.
15 elements · Explore in any order.
1 checklist · 1 heuristic · 3 principles · 10 protocols
15 elements
Bound every autonomous agent loop before it starts
A loop without a hard edge is an outage waiting for a goal condition to fail.
Read the idea ProtocolWrite the completion condition before the agent starts iterating
An agent loop cannot finish reliably if 'done' exists only in the model's mood.
Read the idea ProtocolDetect stagnation separately from completion
More motion is not the same thing as more progress.
Read the idea HeuristicTreat repeated state-action pairs as a loop alarm
If the state did not change, the same action is usually a retry—not a new strategy.
Read the idea PrincipleRetry after a state change; replan after the same failure
A retry is justified by changed conditions; otherwise it is often just repetition.
Read the idea ProtocolCheckpoint meaningful progress before a long agent crosses a fragile boundary
Progress that exists only in the current context window is not durable progress.
Read the idea PrincipleStore durable task state outside the context window
Context is working memory, not a durable project database.
Read the idea ProtocolCompact context by preserving decisions and unresolved work—not every tool result
Compaction should save the project, not the transcript.
Read the idea PrincipleRetrieve context just in time instead of preloading the whole project
Relevant context is valuable; irrelevant context taxes every token after it.
Read the idea ProtocolLet subagents write durable artifacts instead of relaying everything through the coordinator
Every relay is another chance to lose detail.
Read the idea ProtocolEvaluate the same agent task across multiple trials
A nondeterministic system needs more than one audition.
Read the idea ProtocolRead eval transcripts before trusting the score
A score can fail because the agent failed—or because the eval did.
Read the idea ProtocolCombine grader types instead of asking one LLM judge to decide everything
A flexible judge is useful; a monoculture of judgment is fragile.
Read the idea ChecklistTrack tool errors, tool calls, runtime and tokens as agent diagnostics
The path to the answer reveals inefficiency before users notice the bill or latency.
Read the idea ProtocolMonitor the whole agent trajectory, not only individual allowed actions
A safe-looking step can participate in an unsafe-looking trajectory.
Read the idea