fix(release): write release notes to file instead of capturing stdout

[jdx]

Summary

• Use Claude's Write tool to write release notes to a temp file instead of capturing raw stdout with --output-format text, eliminating formatting artifacts (extraneous # in titles, conversational preamble)
• Scope the Write tool to only /tmp/release-notes-output.md using Write($output_file) path restriction
• Upgrade model from claude-opus-4-5 to claude-opus-4-6

Test plan

• Run mise run gen-release-notes v2026.2.9 v2026.2.7 and verify clean output with no preamble or extraneous markdown headers

🤖 Generated with Claude Code

---

Note

Low Risk
Low risk: only adjusts the release-notes generation script flow and CLI invocation, with no impact to runtime behavior beyond how output is captured/validated.

Overview
Switches xtasks/gen-release-notes to have Claude write release notes to a temporary markdown file via the scoped Write($output_file) tool, instead of capturing stdout, to avoid formatting artifacts.

Updates the prompt to instruct file output, upgrades the model to claude-opus-4-6, redirects Claude stdout to /dev/null, and validates the temp file exists and is non-empty before printing it; temp files are cleaned up via a single trap.

^{Written by Cursor Bugbot for commit 1b7e6ba. This will update automatically on new commits. Configure here.}

[gemini-code-assist]

Summary of Changes

Hello @jdx, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request refines the automated release note generation process by leveraging Claude's Write tool to output content directly to a temporary file. This change is designed to produce cleaner, artifact-free release notes by bypassing the need to parse raw standard output. Additionally, the underlying AI model has been updated to a newer version, claude-opus-4-6, to potentially improve generation quality.

Highlights

• Release Note Generation Improvement: Switched from capturing Claude's standard output to using its Write tool to generate release notes directly into a temporary file, which resolves issues with formatting artifacts and conversational preambles.
• Tool Scoping: Scoped the Write tool to specifically allow writing only to /tmp/release-notes-output.md, enhancing security and control over file operations.
• Model Upgrade: Upgraded the Claude model used for release note generation from claude-opus-4-5 to claude-opus-4-6.

Changelog

• xtasks/gen-release-notes
  • Modified the Claude prompt to explicitly instruct the AI to write its output to a specified temporary file.
  • Updated the claude command to use the Write tool with a path restriction and upgraded the model to claude-opus-4-6.
  • Changed the output validation logic to check for the existence and content of the temporary output file instead of capturing stdout.
  • Added cleanup for the temporary output file using trap.

Activity

• No human activity has been recorded on this pull request yet.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

The pull request effectively addresses the issue of formatting artifacts by switching Claude's output mechanism from stdout capture to direct file writing. The changes correctly update the prompt instructions for Claude, modify the command-line arguments for the claude CLI tool to use the Write tool with a scoped path, and adjust the script's logic for temporary file management and output validation. The model upgrade is also applied as described. The implementation is clean and directly achieves the stated goals of the pull request.

[Copilot]

Pull request overview

Updates the release-notes generation task to have Claude write markdown output to a file (via the Write tool) instead of relying on captured stdout, and upgrades the model identifier.

Changes:

• Instruct Claude to write release notes to /tmp/release-notes-output.md and read that file back for output
• Allow the Write tool scoped to the output file path and clean up temp artifacts via trap
• Change the model from claude-opus-4-5-20251101 to claude-opus-4-6

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Using a predictable, fixed filename in /tmp is vulnerable to symlink/hardlink tricks and can lead to clobbering or reading unexpected files. Use a unique temp path (e.g., via mktemp) for output_file and reference that same variable in the prompt so the Write tool is still constrained to exactly that file.

[Copilot]

Using a predictable, fixed filename in /tmp is vulnerable to symlink/hardlink tricks and can lead to clobbering or reading unexpected files. Use a unique temp path (e.g., via mktemp) for output_file and reference that same variable in the prompt so the Write tool is still constrained to exactly that file.

[Copilot]

The output path is duplicated (hard-coded in the prompt and separately in output_file). This can drift if one changes. Prefer constructing the prompt using the output_file variable so there’s a single source of truth.

[Copilot]

The output path is duplicated (hard-coded in the prompt and separately in output_file). This can drift if one changes. Prefer constructing the prompt using the output_file variable so there’s a single source of truth.

[Copilot]

Now that stdout is no longer captured, any unexpected stdout from the claude CLI (preamble, warnings, formatting, etc.) will leak into this script’s stdout and corrupt the intended output. Since the release notes are read from $output_file, redirect Claude’s stdout away (e.g., to /dev/null) and rely solely on the file + captured stderr for diagnostics.

Suggested change

	--allowedTools "Read,Grep,Glob,Write($output_file)" 2>"$stderr_file"; then
	--allowedTools "Read,Grep,Glob,Write($output_file)" >/dev/null 2>"$stderr_file"; then

[github-actions]

Hyperfine Performance

mise x -- echo

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.2.9 x -- echo	22.7 ± 0.6	21.4	28.3	1.00 ± 0.04
mise x -- echo	22.6 ± 0.8	21.3	28.5	1.00

mise env

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.2.9 env	21.4 ± 0.6	20.5	28.8	1.00
mise env	22.1 ± 0.6	20.8	24.2	1.03 ± 0.04

mise hook-env

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.2.9 hook-env	22.2 ± 0.5	21.2	24.0	1.00
mise hook-env	23.0 ± 1.0	21.5	34.7	1.04 ± 0.05

mise ls

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
mise-2026.2.9 ls	20.9 ± 0.5	19.6	23.4	1.00
mise ls	21.3 ± 0.7	19.8	23.6	1.02 ± 0.04

xtasks/test/perf

Command	mise-2026.2.9	mise	Variance
install (cached)	120ms	119ms	+0%
ls (cached)	74ms	73ms	+1%
bin-paths (cached)	78ms	78ms	+0%
task-ls (cached)	536ms	538ms	+0%

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is ON, but a Cloud Agent failed to start.}

[cursor]

mktemp suffix breaks on macOS BSD mktemp

Medium Severity

mktemp /tmp/release-notes-XXXXXX.md uses a template with characters after the trailing Xs, which is a GNU coreutils extension. macOS ships with BSD mktemp, which requires the Xs to be at the very end of the template. On macOS, this command will fail (or worse, create a file literally named release-notes-XXXXXX.md with no randomization). Since the script uses set -euo pipefail, it will abort immediately with a cryptic error on macOS. Dropping the .md suffix (e.g., mktemp /tmp/release-notes-XXXXXX) would be portable across both platforms.