fix: clear nativeArgs when tool-call finalize fails (#1221) - #1634

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

Merge Risk: 🔵 Low · up to 5e913

The fix has integration coverage, but the test can wait for input after its follow-up request. Give it an explicit stopping point before merging.

Security Architecture Review

Security architecture risk: 🔵 Low · up to 5e913

The change improves how incomplete commands are rejected without widening access. An existing exception to that rejection rule and limited coverage of that exception prevent a stronger assurance.

Retained concerns
No architecture-level concerns identified.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

_{Comment @coderabbitai help to get the list of available commands.}

Review status

Thanks for contributing. This comment tracks the review sequence and the next action.

Current step: Address automated review findings and push fixes.

After fixes are pushed and required CI passes, automated review restarts.

Review-state labels are managed by this workflow; do not edit them manually.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

Added an integration-level test alongside the existing simulation-based ones, per the review comment - drives a truncated `write_to_file` call through the real `Task` streaming + `presentAssistantMessage` flow (`recursivelyMakeClineRequests` + a mocked `attemptApiRequest` stream) instead of mirroring the finalize-null logic in an isolated function.

Two things worth noting from building it:

1. `writeToFileTool.handle` legitimately gets called with `partial: true` while the call is still streaming (`BaseTool.handle` short-circuits to a no-op preview hook in that case, per its own doc comment). The test asserts it's never called with `partial: false` specifically - that's the flag that actually reaches `execute()`/disk writes.
2. Confirmed the test fails for the right reason by temporarily reverting the `Task.ts` fix and re-running it: without the fix, `handle` does get called with `partial: false`. An earlier version of this using a `.json` target path was a false negative - the default test task mode's Architect-only markdown file restriction blocked the write independently of the `nativeArgs` guard, masking the actual bug. Switched to a `.md` path so the guard under test is what's actually exercised.

Kept the existing `truncated-native-tool-args.spec.ts` tests too rather than replacing them - they're fast and pin the exact finalize-null branch precisely, complementing the integration test rather than duplicating it.

Full `core/task` suite: 27 files, 382 tests, all passing.

@coderabbitai review

CodeRabbit chat interactions are restricted to organization members for this repository. Ask an organization member to interact with CodeRabbit, or set chat.allow_non_org_members: true in your configuration.

CodeRabbit won't take review commands from non-members here, so it's still holding the CHANGES_REQUESTED from the first commit. What it asked for is in now (an integration test in Task.spec.ts that drives the real streaming flow), along with the params follow-up from taltas's review. Could a maintainer restart the review when there's a chance?

@coderabbitai can you review this please

@taltas I’ll review the current changes in PR #1634, including the new integration test.

Thanks for nudging it, @taltas. It refused because the branch was rebased on the 19th: the commit it reviewed on the 14th (5103e3d) isn't in the PR anymore, so it treats the head as already covered while the old CHANGES_REQUESTED stays. Its own suggestion should do it, the full review command instead of review. Dismissing that stale review from the 14th would unblock it too, if that's easier.

@coderabbitai please dismiss your review, its stale

#!/bin/bash
set -euo pipefail
gh api repos/Zoo-Code-Org/Zoo-Code/pulls/1634 --jq '{number, state, head_sha:.head.sha, updated_at}'
gh api repos/Zoo-Code-Org/Zoo-Code/pulls/1634/reviews --paginate --jq '.[] | {id, user:.user.login, state, commit_id, submitted_at, body, html_url}'

@taltas, I checked the review records. The September 14 CHANGES_REQUESTED review is attached to an older commit and appears stale. A second CodeRabbit CHANGES_REQUESTED review was submitted on September 26 against the current head commit, so dismissing the older review will not clear the current review request.

Dismissal changes the PR’s review state. Please confirm that you want me to dismiss only the September 14 review.

You are interacting with an AI system.

@coderabbitai dismiss all reviews please

#!/bin/bash
set -euo pipefail
gh api repos/Zoo-Code-Org/Zoo-Code/pulls/1634 --jq '{number,state,head_sha:.head.sha}'
gh api repos/Zoo-Code-Org/Zoo-Code/pulls/1634/reviews --paginate --jq '.[] | {id,user:.user.login,state,commit_id,submitted_at,html_url}'

#!/bin/bash
set -u
repo="repos/Zoo-Code-Org/Zoo-Code/pulls/1634"
for id in 5194833057 5324086215; do
  echo "Dismissing CodeRabbit review $id"
  if gh api -X PUT "$repo/reviews/$id/dismiss" -f message="Dismissed at the request of `@taltas` on PR `#1634`." --jq '{id,state,user:.user.login,html_url}'; then
    echo "Dismiss request succeeded for $id"
  else
    echo "Dismiss request failed for $id" >&2
  fi
done
echo "Current CodeRabbit review states:"
gh api "$repo/reviews" --paginate --jq '.[] | select(.user.login == "coderabbitai[bot]") | {id,state,commit_id,html_url}'

@taltas, I tried to dismiss both CodeRabbit reviews on PR #1634. GitHub returned 404 Not Found for each request. I checked their states afterward; both remain CHANGES_REQUESTED.

Please dismiss the two reviews in the GitHub UI, or retry the request after API access is available. I did not change anyone else’s reviews.

You are interacting with an AI system.