Skip to main content
Use the Query skill before non-trivial work so your agent can align with existing project decisions, patterns, and gotchas. The normal workflow is agent-driven: ask the agent to check ByteRover, then let it use the retrieved memory while it works.
Run Query from inside the project folder. If the project should use a named space, bind it first with Bind.

​
Ask your agent to query

Use natural language. Include the area of the codebase or decision you are about to touch.
Other examples:
The agent should run the query from inside the project folder so ByteRover can resolve the correct space.

​
Review the answer

The agent should summarize the relevant memory and show the source when ByteRover returns one. Use the answer as project context, not as a blind command. If the recorded memory conflicts with the current task, ask the agent to explain the conflict before changing direction.

​
Narrow the query

If the answer is too broad, ask a more specific follow-up:
If the answer misses a topic you expected, mention the likely area:

​
Record what is missing

If nothing useful appears, continue the work. When the decision becomes clear, ask the agent to record it:

​
Manual reference

Most users do not need to run the script directly. The connected agent runs it from the installed ByteRover skill directory. The script shape is:
The command returns JSON with:

​
Read a full topic

When a hit looks useful, read the full topic:
The agent should treat recorded decisions and rules as constraints unless it explains why the current task needs to diverge.

​
Next steps

Record

Save durable project knowledge after work is complete.

Dream

Review cleanup and consolidation proposals for saved memory.