[NikitaMGrimm]

Problem

An attached thread has two IDs: the context pairing ID shown in the marker and link (thread_<uuid>), and the canonical thread ID. The generated instruction only says t3_thread_read(threadId), so agents copy the pairing ID and get thread_not_found even though the thread exists. Reproduced on a live Windows nightly: the pairing ID failed, and the canonical ID read the same thread.

Change

The shared provider context formatter now writes the call with its argument, for example t3_thread_read({"threadId":"<canonical id>"}), using record.threadId. Marker/envelope pairing, exact thread lookup, and permissions are unchanged.

Scope and approval

Related: #15849 and its duplicate #19639, both closed without a fix. The triage on #15849 lists a prompt-side correction as one possible direction. This is submitted under the small, focused obvious-bug exception: one generated instruction now supplies the existing tool's correct argument, with no ID aliasing, permission change, or new workflow. It does not implement the server-side resolver options from those reports, and it does not fix histories that were already injected. #22452 fixed a different problem (a missing mcp: prefix).

Verification

• vp test run --config ../../vite.config.ts src/composerContextReferences.test.ts in packages/shared: the existing projection case now asserts the exact t3_thread_read({"threadId":"abc"}) call. It fails against the unchanged formatter and passes 17/17 with the fix.
• Against the live reproduction, calling t3_thread_read with the generated argument returned the existing thread. No live state was modified.
• Changed-file format and lint pass. An independent review of the fix (before the test was slimmed) found no actionable issues.

This applies to newly projected context on the provider start and steer paths shared by all providers and clients. There is no visible UI change.

Note

🤖 Agent assistance via T3 Code

• GPT 6.1 Sol via Codex: implementation; review (independent code and requirement review); supporting work (investigation and verification).
• Claude Opus 5.5 via Claude Code: supporting work (test slimming and PR writing).

[coderabbitai]

Warning

Review limit reached

Only 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 5 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used all 10 included reviews currently available.

Learn how review limits work.

Review configuration:

Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: main
• File: .coderabbit.config.ts
• Commit: f88b33a
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Advanced
• Run ID: ebecd8cd-86d5-4242-b181-9c9f7b0ed8ce

Commits

Reviewing files that changed from the base of the PR and between dce4af3 and b4e9cea.

Files selected for processing: 2

• packages/shared/src/composerContextReferences.test.ts
• packages/shared/src/composerContextReferences.ts

No actionable comments were generated in the recent review.

ℹ️ Recent review info

⚙️ Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: main
• File: .coderabbit.config.ts
• Commit: 22840ab
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Advanced
• Run ID: 07157c31-b1ac-4683-aefe-19ad8d806cb5

📥 Commits

Reviewing files that changed from the base of the PR and between 1b39138 and dce4af3.

📒 Files selected for processing: 1

• packages/shared/src/composerContextReferences.test.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 9 remain after this review.

Walkthrough

The attached-thread context payload now includes the record’s thread ID in the generated t3_thread_read call. The projection test checks the serialized argument.

Changes: Attached Thread Read

Layer / File(s)	Summary
Serialize thread ID in read call packages/shared/src/composerContextReferences.ts, packages/shared/src/composerContextReferences.test.ts	The payload uses the record’s thread ID as the JSON argument to t3_thread_read. The projection test checks the expected serialized argument.

Priority: Normal

Estimated code review effort: 2 (Simple) | ~8 minutes

Change: Bug fix

Possibly related issues

• Claude thread permanently wedged after MCP registration timeout: stale pending handoff forces fresh session, full-thread handoff exceeds budget, /compact can't recover #22198 proposes a t3_thread_read pointer for thread handoffs. This change supplies the attached thread’s ID to generated read calls, but does not address the stale delivery marker or handoff budget failure.

Suggested reviewers: t3dotgg

Merge Risk: Minimal · up to dce4a

The changed attached-thread projection uses the valid canonical ID, with no merge-blocking risk established.

Pre-merge checks: 4

Check name	Status	Explanation
Title check		The title clearly and concisely describes the main fix: attached threads now use a canonical read call.
Description check		The description includes complete Problem, Change, Scope and approval, and Verification sections. It explains the failure, implementation, scope limits, related issues, test commands, observed results…
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.

Full details: Description check

Explanation

The description includes complete Problem, Change, Scope and approval, and Verification sections. It explains the failure, implementation, scope limits, related issues, test commands, observed results, and agent assistance.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

[maria-rcks]

Note

Written by gpt-6.1-sol on behalf of Maria

kept @NikitaMGrimm's implementation unchanged and merged current main into the writable contributor branch. the final diff remains two changed lines: serialize the canonical record.threadId in the read call and assert that call in the existing projection test. nothing needed dropping or adding.

verification on b4e9ceafae3497b1ddc3934356887ab5c98315e0:

• reproduced the regression with main's formatter: exactly the attached-thread assertion failed, with 16 other tests passing.
• restored the fix: all 17 tests passed.
• blacksmith vp run check:changed github/main passed format, lint, both knip passes, typechecks for 29 affected packages, and the focused tests.
• two independent code reviews approved this head with no findings; start and steer share the corrected projector and the argument matches the tool schema.

unverified: a fresh live provider/client attachment-to-read round trip. there is no visible ui change. github ci passed, including build/typecheck, lint, web tests, all eight server shards, the shared tests, and release smoke. github reports MERGEABLE / CLEAN. coderabbit reports a passing check but skipped its fresh review because the review budget is exhausted; the prior implementation review had no findings. no unresolved review threads remain.