Read tool status message omits `MAX_BYTES` cap when `MAX_LINES` is also hit

[jiang1997]

What version of Kimi Code is running?

0.3.0

Which open platform/subscription were you using?

Kimi Code

Which model were you using?

No response

What platform is your computer?

Debian(WSL)

What issue are you seeing?

When reading a file that exceeds both the MAX_LINES and MAX_BYTES_KB limits, the <system> status tag in the tool result only reports the line limit. The byte limit is silently omitted even though it was also triggered.

For example, the status might say:

<system>... Max 1000 lines reached.</system>

But it does not mention that MAX_BYTES was also hit, which could mislead the model into thinking only the line cap mattered.

What steps can reproduce the bug?

1. Prepare a file that exceeds both MAX_LINES (1000) and MAX_BYTES_KB (100 KB) limits. For example, a file with 1001 lines where each line is very long (e.g., 120+ characters).

2. Use the Read tool to read this file.

3. Inspect the wire.jsonl file — in the tool result's <system> status tag, you will see:

   <system>... Max 1000 lines reached.</system>
   

   But not the Max 102400 bytes reached. message, even though the byte limit was also hit.

The root cause is in packages/agent-core/src/tools/builtin/file/read.ts: the limit checks are chained with else if, so when both maxLinesReached and maxBytesReached are true, only the first branch executes.

What is the expected behavior?

Both limits should be reported independently when both are triggered:

<system>... Max 1000 lines reached. Max 102400 bytes reached.</system>

The EOF message ("End of file reached.") should only appear when neither truncation limit was hit.

Additional information

The fix is straightforward — change else if to independent if statements and tighten the EOF condition. I have a branch ready and can open a PR once this issue is triaged.

[jiang1997]

@caigee-cmd Thanks for the PR.

I opened #94 and explicitly mentioned that I already had a branch ready and was waiting for triage before opening a PR.

Even though this is a small fix, please coordinate on the issue first in cases where someone has already stated they are actively working on the implementation. Otherwise it creates duplicated work.

[caigee-cmd]

@jiang1997 You're completely right, and I apologize. You clearly stated in the issue that you had a branch ready and were waiting for triage before opening a PR — I should have coordinated here first instead of jumping straight to a PR. That created duplicated work, which is on me.

I'm closing #99 so you can open yours as intended. Sorry for stepping on it, and thanks for raising this directly.