What does this skill do, and when should you use it?
Session Code Review examines the files changed in the current session and reports concrete, actionable findings as inline code comments. It compares the changed ranges with the surrounding codebase while checking correctness, edge cases, regressions, error handling, security, consistency, and newly introduced test or documentation gaps. Each finding is attached to an exact file URI and line range rather than summarized at file level. The skill is review-only: it does not modify files or perform commits or pushes.
Determines the current session's changed files, using commands such as git status or git diff; reads relevant ranges from each changed file and compares them with the rest of the codebase; checks correctness, edge cases, bugs, regressions, error handling, security and data handling, naming and consistency, and missing tests or documentation; calls the addComment tool for each issue with the exact file URI and line range; favors fewer, higher-signal findings and stops after every changed file has been reviewed.
- A developer clicks Run Code Review in the Changes toolbar and wants the current session's edits checked immediately.
- A maintainer reviews a multi-file change for regressions or edge cases and wants findings placed beside the relevant code.
- A security reviewer checks changed code for security or data-handling issues that need precise location-based feedback.
- A team performs a pre-merge review for missing error handling, tests, or documentation without allowing the reviewer to alter the worktree.
How do you install this skill?
- Posting addComment comments changes external state, but the workflow does not require per-comment or final user confirmation or explain how mistaken comments can be withdrawn.
- The skill depends on the current session's change set and the addComment tool without specifying diagnosable behavior for missing tools, empty changes, failures, or unlocatable ranges.
- It does not state how to avoid copying secrets, personal data, or other sensitive content into comments, nor does it document Chinese-language or mainland-China environment support.
- Shell / CLI
- Local filesystem
GitaddComment tool
The source does not provide a standalone installation command or client-specific installation steps. The skill is located at src/vs/sessions/skills/code-review/SKILL.md in microsoft/vscode. The repository bundles 61 separate skills, so installation should follow the procedure provided by the Agent Skills client for that collection.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/microsoft/vscode.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/src/vs/sessions/skills/code-review" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
In a session containing code changes, click Run Code Review in the Changes toolbar. The skill reviews the session's changed files and posts findings with addComment against exact files and line ranges. It does not edit files, commit, or push changes. The source does not document the behavior when there are no changed files.
What are this skill's strengths and limitations?
- Covers correctness, regressions, security, error handling, consistency, testing, and documentation gaps.
- Requires findings to target an exact file URI and line range, making them easier to act on.
- Explicitly restricts itself to review comments and forbids file edits, commits, and pushes.
- Prioritizes a small number of high-signal findings over minor stylistic issues.
- Its scope is the current session's changed files, not a general full-repository audit.
- It depends on shell access, filesystem reads, Git commands, and the addComment tool.
- The source provides no standalone installation procedure, test suite, or client-by-client validation evidence.
- The source does not specify recovery behavior for no changes, unreadable code, or addComment failures.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Session Code Review this page ✓ Microsoft · Official | 51 · Use with care | ★ 194k | 11d ago | MIT |
| Open Code Review Skill | 58 · Recommended | ★ 47k | 11d ago | Apache-2.0 |
| Review Agent Code Review Skill ✓ OpenAI · Official | 52 · Use with care | ★ 129k | 11d ago | Apache-2.0 |
| Open Code Review Delegated Review | 58 · Recommended | ★ 47k | 11d ago | Apache-2.0 |
| CodeRabbit Code Review | 57 · Use with care | ★ 193 | 10d ago | MIT |
How did FollowSkills review this skill?
The skill narrowly scopes work to reviewing current-session changes, posting line-level comments, and forbidding file edits, commits, and pushes, which supports least privilege and scope control. It does not require user confirmation before publishing comments, disclose comment data flow or sensitive-data handling, or specify rollback and recovery, so points are deducted.
The workflow is internally coherent across change discovery, per-file review, and per-issue commenting. However, it does not define a reliable current-session change source, the availability or contract of addComment, handling for abnormal input, or diagnosable failure feedback. Static calibration also caps this score at 10.
The name, description, and trigger are clear for users invoking Run Code Review from the Changes toolbar. Non-fit boundaries, input limits, output-format details, and Chinese-language support are not stated, leaving environment and semantic-trigger evidence incomplete.
The file has basic front matter, a clear title, and a progressive workflow. Repository material supplies MIT licensing, Microsoft attribution, version information, and maintenance signals. The skill itself lacks versioning, changelog, installation or dependency notes, FAQs, known limitations, and an explicit update path, so points are deducted.
The goal and output are concrete: produce specific, actionable inline comments for each issue while avoiding low-signal stylistic nitpicks. No representative output, quality verification, or comparative-benefit evidence is included, so only limited static effectiveness credit is justified.
The skill text is auditable, and repository files provide licensing, version, and some CI evidence. There are no skill-specific tests, an explicit addComment contract, or independent reproduction records covering the key path, so verifiability remains limited.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →