How to Reduce Repetitive Work and Unnecessary Usage?

Write Success Criteria at the Start

First specify what needs to be delivered, which scopes to check, and under what results the task can be concluded. Compared to continuously adding "keep digging deeper," clear end conditions make it easier to control workload. For example, require locating one error cause first, then decide whether to check other modules.

Provide relevant materials only, avoiding pasting the same large background text in every round.

Save and Reuse Existing Results

For long tasks, have the Agent save outputs as files and later read only relevant fragments; if interrupted, first check these results instead of repeating all commands from scratch.

For example, when investigating a startup error, first review the current configuration and latest error logs, and only expand checks if evidence points to dependency issues. Writing "first determine if the next step is needed" into the request can reduce aimless tool installations, repeated searches, and overly long reports.

1. Complete a small sample or lightweight check first.
2. Decide whether to expand the scope based on findings.
3. Clearly document verified facts, output locations, and pending tasks.
4. For new issues, only add changed conditions.

Choose a Processing Method Proportional to the Task

For daily organization, start with the current default model; use other available options for complex problems. When querying the web, clearly specify the topic and time range to avoid many irrelevant results entering the analysis. Reducing usage should not rely on omitting necessary verification; prioritize reducing repetition and deviation, and observe actual usage records to identify which tasks are most worth adjusting.