[CNSeniorious000]

Skill and MCP display names also leak into ChatML history, replacing the tool name Codex actually called. For example, a call recorded as exec_command appears as skill:bug-mentor in the next generation's input.

Use the recorded tool name for assistant tool calls and tool replies. Keep observation names and trace tags unchanged.

Tests

• Cover skill calls with skill_tags on and off, plus MCP calls; assert original names in both history roles and unchanged observation names.
• pnpm run lint and pnpm test pass (127 tests).
• Bundled Stop hook → local OTLP receiver: verify exported history, display names, sidecar writes, and repeat-Stop deduplication. The history assertion fails on unmodified v0.4.3.

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[CLAassistant]

All committers have signed the CLA.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-06T03:19:53.286734Z	8e01bc0	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[sangkyoonnam]

The two history fields lose the "tool" fallback for an empty or missing name. The parent builds both through toolObservationName(tc), whose final fallback (parent trace.ts:284) is tc.name || "tool". The head assigns tc.name directly in both places.

I ran pnpm --config.manage-package-manager-versions=false exec vitest run plugins/tracing/test/trace.test.ts -t "keeps real tool names|captures web search, local shell, and MCP" 3 times on the head (3 passed each) and 3 times with only trace.ts reverted to its parent version (3 failed each). The two inline suggestions keep the fallback.