Adds skill for generating engine diffs for new releases.

[gaaclarke]

Adds a skill for generating a "what's new" for the flutter engine across branches. This skill wasn't able to generate the information directly needed for a blog post but it was informative to writing highlights for a blog post. It's probably worth checking this in and trying to refine it further in the future.

Pre-launch Checklist

• I read the Contributor Guide and followed the process outlined there for submitting PRs.
• I read the AI contribution guidelines and understand my responsibilities, or I am not using AI tools.
• I read the Tree Hygiene wiki page, which explains my responsibilities.
• I read and followed the Flutter Style Guide, including Features we expect every widget to implement.
• I signed the CLA.
• I listed at least one issue that this PR fixes in the description above.
• I updated/added relevant documentation (doc comments with ///).
• I added new tests to check the change I am making, or this PR is test-exempt.
• I followed the breaking change policy and added Data Driven Fixes where supported.
• All existing and new tests are passing.

If you need help, consider asking for advice on the #hackers-new channel on Discord.

If this change needs to override an active code freeze, provide a comment explaining why. The code freeze workflow can be overridden by code reviewers. See pinned issues for any active code freezes with guidance.

Note: The Flutter team is currently trialing the use of Gemini Code Assist for GitHub. Comments from the gemini-code-assist bot should not be taken as authoritative feedback from the Flutter team. If you find its comments useful you can update your code accordingly, but if you are unsure or disagree with the feedback, please feel free to wait for a Flutter team member's review for guidance on which automated comments should be addressed.

[gaaclarke]

@reidbaker how do you like including this and iterating on it?

[reidbaker]

If it is valuable to you we can add it. Let me give it a review

[reidbaker]

Let's be crisp about the formats you want to accept here. Are they tags?

Also I think for now given how infrequently we expect this skill to be used we should say that the skills should be run only when called out by name.

If you want to leave it with the prompts I would ask that you move the skill to a agent for the engine team "engine-agent"

[gaaclarke]

Let's be crisp about the formats you want to accept here. Are they tags?

That's part of the value I see this skill providing. There is a lot of knowledge that goes into determining a git ref from a semantic version name.

Also I think for now given how infrequently we expect this skill to be used we should say that the skills should be run only when called out by name.

Sounds good, done.

[reidbaker]

This summary section is both common when you ask Gemini based models to author a skills and also un-needed token bloat. Information here needs to be in the description or deleted.

[gaaclarke]

Done

[reidbaker]

Non blocking but I would encourage you to author this not as a script but as a dart package that has tests. Especially given the complexity

[gaaclarke]

I want the skill and the script to evolve together. Maybe it will evolve into a script that will have independent value at some point.

[reidbaker]

If your script is doing the bulk of the work and not the llm running the skill is the script alone valuable enough to check on without the skill?

[gaaclarke]

The llm provides 2 valuable features:

1. Mapping semantic names to git references, example "version 3.47" means 3.47.0-0.1.pre today. It will be something different in the future.
2. Summarizes and characterizes the changes outlined by the script. The script is there just to provide a deterministic base for it to summarize from.

[reidbaker]

I would expect the llm to be used more in this skill to help with the identification and summarization

[gaaclarke]

It is providing summarizations, I wanted it to generate a paper trail too though so that it's auditable.

The way I used it was to read the summarization, then look through the generated files to get more information. Wether the summary was helpful was hit or miss and I hope we can refine it.

[gemini-code-assist]

Code Review

This pull request introduces a new agent skill and a helper Dart script to generate release summaries and diff files for the Flutter engine. Feedback on the implementation highlights a critical correctness issue where the script incorrectly assumes the engine source code resides within the main Flutter repository. Additionally, suggestions were made to document public members per the style guide, use a Set to eliminate duplicate git ref candidates, and refine the commit categorization logic to prevent false positives from substring matches.

[gemini-code-assist]

The class CommitInfo (and other public members in this file, such as ReleaseAnalysis, findRepoRoot, runGit, etc.) lacks documentation comments. According to the repository style guide, all public members should have documentation using ///.

References

1. All public members should have documentation. ^(link)

[gemini-code-assist]

The list of candidates contains duplicate entries when cleanVersion starts with '3.' (e.g., 'origin/flutter-3.47-candidate.0' and 'flutter-3.47-candidate.0' are added twice). Using a Set<String> instead of a List<String> is a cleaner and more efficient way to automatically eliminate duplicates.

Suggested change

	final candidates = <String>[
	cleanVersion,
	if (!cleanVersion.contains('.')) '3.$cleanVersion.0',
	if (cleanVersion.startsWith('3.') &&
	!cleanVersion.contains('-') &&
	cleanVersion.split('.').length == 2)
	'$cleanVersion.0',
	'origin/flutter-$cleanVersion-candidate.0',
	'flutter-$cleanVersion-candidate.0',
	if (cleanVersion.startsWith('3.'))
	'origin/flutter-${cleanVersion.split('.').take(2).join('.')}-candidate.0',
	if (cleanVersion.startsWith('3.'))
	'flutter-${cleanVersion.split('.').take(2).join('.')}-candidate.0',
	'v$cleanVersion',
	'v$cleanVersion.0',
	];
	final Set<String> candidates = <String>{
	cleanVersion,
	if (!cleanVersion.contains('.')) '3.$cleanVersion.0',
	if (cleanVersion.startsWith('3.') &&
	!cleanVersion.contains('-') &&
	cleanVersion.split('.').length == 2)
	'$cleanVersion.0',
	'origin/flutter-$cleanVersion-candidate.0',
	'flutter-$cleanVersion-candidate.0',
	if (cleanVersion.startsWith('3.'))
	'origin/flutter-${cleanVersion.split('.').take(2).join('.')}-candidate.0',
	if (cleanVersion.startsWith('3.'))
	'flutter-${cleanVersion.split('.').take(2).join('.')}-candidate.0',
	'v$cleanVersion',
	'v$cleanVersion.0',
	};

[gemini-code-assist]

The categorizeCommit function has several redundancies and potential false positives:

1. Redundancies:

   • title.startsWith('Roll Skia') etc. are redundant because lower.contains('roll skia') is already checked.
   • lower.contains('[android]') is redundant because lower.contains('android') is checked.
   • Same for [ios], [macos], [darwin], [windows], [linux].

2. False Positives:

   • lower.contains('flow') matches "workflow" (very common in CI/tooling), causing CI commits to be miscategorized under "Impeller & Graphics Rendering".
   • lower.contains('format') matches "information" and "transformation", causing general commits to be miscategorized under "Build System, CI & Tooling".

Using RegExp with word boundaries (\b) resolves these false positives elegantly.

String categorizeCommit(String title) {
  final String lower = title.toLowerCase();

  if (lower.contains('roll skia') ||
      lower.contains('roll dart sdk') ||
      lower.contains('roll icu') ||
      lower.contains('roll harfbuzz') ||
      lower.contains('roll angle')) {
    return '🔄 Dependency Rolls';
  }

  if (lower.contains('impeller') ||
      lower.contains('ubersdf') ||
      lower.contains('flutter gpu') ||
      lower.contains('display_list') ||
      lower.contains('displaylist') ||
      lower.contains('vulkan') ||
      lower.contains('metal') ||
      lower.contains('opengl') ||
      lower.contains('shader') ||
      lower.contains('render') ||
      RegExp(r'\bflow\b').hasMatch(lower)) {
    return '🚀 Impeller & Graphics Rendering';
  }

  if (lower.contains('web_ui') ||
      lower.contains('web_sdk') ||
      lower.contains('wasm') ||
      lower.contains('skwasm') ||
      lower.contains('html') ||
      lower.contains('canvaskit') ||
      RegExp(r'\bweb\b').hasMatch(lower)) {
    return '🌐 Web Engine & Wasm';
  }

  if (lower.contains('android') ||
      lower.contains('agp') ||
      lower.contains('gradle') ||
      lower.contains('embedding/engine')) {
    return '📱 Android Embedding';
  }

  if (lower.contains('darwin') ||
      lower.contains('ios') ||
      lower.contains('macos') ||
      lower.contains('xcode') ||
      lower.contains('metalview')) {
    return '🍎 iOS & macOS Embeddings';
  }

  if (lower.contains('windows') ||
      lower.contains('linux') ||
      lower.contains('win32') ||
      lower.contains('embedder')) {
    return '🪟 Windows & Linux Desktop Embeddings';
  }

  if (lower.contains('semantics') ||
      lower.contains('accessibility') ||
      lower.contains('typography') ||
      lower.contains('txt') ||
      lower.contains('font') ||
      lower.contains('text input') ||
      lower.contains('autofill') ||
      RegExp(r'\ba11y\b').hasMatch(lower)) {
    return '🔤 Text, Typography & Accessibility';
  }

  if (lower.contains('build.gn') ||
      lower.contains('tools') ||
      lower.contains('testing') ||
      lower.contains('header_guard') ||
      lower.contains('license') ||
      RegExp(r'\bci\b').hasMatch(lower) ||
      RegExp(r'\bformat\b').hasMatch(lower)) {
    return '🛠️ Build System, CI & Tooling';
  }

  return '⚙️ Core Runtime & Shell';
}

[reidbaker-agent]

whoops i was logged into the wrong account

[reidbaker]

I still think the repo would be better off with the script as a package but approving to unblock. Thank you for responding to my feedback.