aiGatekeeper (VS Code Extension)

Intelligent, Git-Aware Code Reviewer for Visual Studio Code powered by GitHub Copilot. Detects files that were Added, Modified, Renamed, or Deleted in your Git working tree or branch diff, reviews the changes using the official VS Code Language Model API, and generates PR-style inline reviewer comments called Review ToDo at the exact code locations requiring attention.

Key Features

Reviews Changes, Not Entire Repositories: Scopes analysis strictly to Git hunks, Added files, Modified lines, Renamed assets, and Deleted modules.

PR-Style Review ToDo Comments: Actionable reviewer findings structured as:

Review ToDo
────────────────────────
[HIGH] Potential null reference

This change can cause `user.profile` to be accessed
when `user` is undefined.

Fix:
Add a null check before accessing profile.

[Suggest Fix] [Preview Fix] [Ignore]

Diff-Aware Location Validation: Findings are strictly verified against actual Git diff hunks. If an AI finding references an unchanged line, it is safely re-anchored to the nearest changed line or designated as file-level. Never generates comments against unrelated code.
VS Code Language Model API: Native, direct integration with GitHub Copilot models (gpt-4o, gpt-4o-mini, gpt-4).
Comprehensive Review Modes:
- Review Working Tree (staged + unstaged + untracked)
- Review Staged Changes
- Review Unstaged Changes
- Review Branch (baseBranch...HEAD)
- Review Current File Changes
Native Editor Integrations:
- VS Code Diagnostics collection (ai-review-todo) with severity mapping
- Hover tooltips with action links
- CodeLens indicators above changed blocks
- Quick Fix / Code Actions menu integration
- Interactive Tree View in the Activity Bar
- Dedicated Webview Review Dashboard
Safe AI Fixes:
- Validates code boundaries and prevents directory traversal
- Verifies that file content has not drifted before applying
- Side-by-side diff preview dialog
- Applied via native vscode.WorkspaceEdit with standard VS Code Undo support
Automated Credential Redaction: Scans and redacts API keys, tokens, and private keys prior to model submission.
Exporting: One-click export of findings to Markdown and JSON formats for PR descriptions and team handoffs.

Review Categories

Bugs: Logic errors, null/undefined dereferences, incorrect async/state handling, race conditions, boundary bugs.
Security: SQL/command injection, path traversal, hardcoded secrets, authorization bypasses, sensitive data leakage.
Performance: Unnecessary expensive operations, n+1 queries, memory leaks, blocking computations.
Reliability: Missing error handling, unsafe recovery, timeout handling, unhandled rejections.
Maintainability: Duplication, excessive complexity, tight coupling, breaking modularity.
Testing: Missing tests, broken test assumptions, weak assertions.
Compatibility: Breaking API changes, broken imports from deleted files, type mismatches.
Code Quality: High-signal issues only — noisy formatting or trivial naming preferences are ignored.

Extension Settings

Setting	Type	Default	Description
`aiReviewToDo.minimumSeverity`	`string`	`"LOW"`	Minimum severity threshold (`CRITICAL`, `HIGH`, `MEDIUM`, `LOW`, `INFO`).
`aiReviewToDo.reviewCategories`	`array`	All	Categories evaluated during review.
`aiReviewToDo.includeUntracked`	`boolean`	`true`	Include untracked files in working tree review.
`aiReviewToDo.maxFiles`	`number`	`40`	Maximum files to analyze per review session.
`aiReviewToDo.maxFileSize`	`number`	`500`	Maximum file size in KB before skipping.
`aiReviewToDo.maxTotalContext`	`number`	`32000`	Maximum character budget for project context.
`aiReviewToDo.excludePatterns`	`array`	`[...]`	Glob exclusion patterns.
`aiReviewToDo.showDiagnostics`	`boolean`	`true`	Show inline editor diagnostics.
`aiReviewToDo.enableFixSuggestions`	`boolean`	`true`	Allow generating code fixes via Copilot.
`aiReviewToDo.requireFixConfirmation`	`boolean`	`true`	Require confirmation before applying edits.
`aiReviewToDo.telemetryEnabled`	`boolean`	`false`	Disabled by default.
`aiReviewToDo.modelFamily`	`string`	`"gpt-4o"`	Copilot model family to request.

Development & Packaging

# Install dependencies
npm install

# Compile TypeScript
npm run compile

# Package extension into .vsix file
npm run package
# Generates ai-review-todo-1.0.0.vsix

Running and Debugging in VS Code

Open this repository in Visual Studio Code (code .).
Press F5 or select Run Extension from the Run & Debug view.
A new Extension Development Host window will launch with AI Review ToDo active.
Open any Git repository and run AI Review ToDo: Review Changes from the Command Palette (Ctrl+Shift+P / Cmd+Shift+P).

License

MIT

aiGatekeeper

Vishal Sharaf

aiGatekeeper (VS Code Extension)

Key Features

Review Categories

Extension Settings

Development & Packaging

Running and Debugging in VS Code

License