Repository navigation
Conversation
Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>
|
Warning Review limit reachedOnly developers with an assigned seat can use this organization's usage-based review budget, and seats here are assigned manually. Ask an admin to assign a seat, or change the review continuation mode in Billing. Next included review available in 13 minutes. View limit details
Comment |
ApprovabilityVerdict: Approved at Macroscope's review found this PR approvable — This is a small, self-contained web composer bug fix that preserves thread context records through copying and stashing, then clears them consistently from the draft. Its runtime impact is limited to those existing user actions, with no schema, deployment, security, billing, default, or lint-policy changes. You can add or adjust custom eligibility rules. Learn more. |
Problem
Attach a thread to a prompt with
@, then copy that chip into another draft, or stash the prompt and restore it in another thread on the same environment. Both paths export terminal, review-comment and preview context records but not thread records. Only thet3-contextlink survives, so the destination cannot resolve the chip and sends it as unavailable context instead of the referenced thread. Stashing also leaves the thread records behind in the source draft after clearing its prompt.Change
The importer already handled thread records; it just never received them.
Scope and approval
A small fix for an obvious bug (6 lines), so it uses the no-prior-issue exception.
Verification
main, 3 of 4 fail: clipboard preservation, stash preservation and source cleanup (log). With the fix all 4 pass (log): the thread ID survives a same-environment import alongside a terminal context, only selected records are copied, and a rejected stash write keeps the source records. The probes evaluate extracted component code, so they are not committed. Existing draft, stash and context-reference tests pass (214, log).vp lintandapps/webtypecheck: clean.Found by a GPT-6.1 Sol (high) audit and implemented by GPT-6.1 Sol (medium) in T3 Code. Reviewed by Claude Opus 5.5. Coordinated by Claude Opus 5.5 in Claude Code via T3 Code. Agent-authored, sponsored by Alex Southwell.
🤖 Generated with Claude Code