- Home
- Skills
- Code Quality & Review
- Smart Git Diffs
Works with the AI Agents you already use
Smart Git Diffs
Summarize large git diffs and branch changes into Markdown using a CLI with map-reduce and secret redaction.
Free
Smart Git Diffs
Example session with this skill installed
Summarize the changes between v1.2.0 and HEAD, but exclude the documentation folder and redact any secrets found.
- Read your context and instructions
- Compiled the smart git diffs
I have generated a summary of the changes between v1.2.0 and HEAD. I used --exclude-default-noise and --redact-secrets to ensure the output is clean and secure.
Git Change Summary: v1.2.0 -> HEAD
Overview
This range includes 14 commits focusing on the core engine refactor and the addition of the new plugin architecture.
Key Changes
- Core Engine: Migrated the event loop to a more efficient priority queue implementation in
src/engine/. - Plugin System: Introduced
src/plugins/base-plugin.tsto allow third-party extensions. - Configuration: Updated the schema to support nested options for provider-specific settings.
- Bug Fixes: Resolved a race condition in the file watcher and improved error handling for network timeouts.
File Statistics
- Modified: 8 files
- Created: 3 files
- Deleted: 0 files
[!NOTE]
Secrets detection was active. Any strings matching credential patterns were masked with[REDACTED]before processing.
Next steps
- Review the
src/plugins/directory for implementation details. - Run the test suite to ensure the engine refactor hasn't introduced regressions.
- Draft the final release notes using this summary as a base.
Connects securely to your tools. The creator never sees your data.
About this skill
Git diffs often exceed context windows or contain too much noise for a standard agent to process reliably. This skill provides a specialized interface for the smart-diff CLI, enabling you to generate structured Markdown summaries of any git range while managing token limits and sensitive data. It handles the heavy lifting of diff reduction and batching so your agent can focus on high-level reasoning.
What it does
- Smart Summarization generates structured Markdown summaries of changes between any two git refs or branches.
- Token Management uses map-reduce logic to process large diffs that exceed standard LLM context windows.
- Noise Filtering automatically excludes lockfiles, build artifacts, and dependency directories to focus on source code.
- Secret Redaction masks credentials and sensitive patterns before data leaves the local environment.
- Commit Filtering allows targeting specific types of changes using regex for feature or fix commits.
How it works
- Environment Check verifies Node.js version 22.22.1+ and the presence of LLM provider credentials in the environment.
- Range Selection identifies the correct git refs, such as
main...HEADor specific tags, based on user intent. - Execution runs the CLI via
npxwith flags for secret redaction, noise exclusion, and map-reduce if the diff is massive. - Reporting outputs the raw Markdown summary directly from the CLI to ensure fidelity.
Frameworks & tools
Requires Node.js 22.22.1 or newer. Works with OpenAI, Anthropic, Google Gemini, and AWS Bedrock via environment variables.
Why this beats prompting it yourself
Prompting an agent to read a raw diff often results in truncated output or missed details when files are large. This skill uses a dedicated CLI with built-in token reduction and secret masking, ensuring consistent results even for enterprise-scale changes.
Use cases
- Generate professional pull request descriptions based on branch changes.
- Draft release notes from git history between two tags.
- Review large architectural shifts using map-reduce batching.
- Audit specific changes by filtering for
featorfixcommit patterns.
Known limitations
Requires a local Node.js environment and existing git repository. Summarization quality depends on the external LLM provider configured in environment variables.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.
Requires Node.js 22.22.1+. Only works inside a git repository. Must have LLM provider credentials (e.g., OPENAI_API_KEY) pre-set in the environment.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 11 days ago
- Free with an Agensi account
- Accounts
OpenAI
Anthropic
Groq
Google AILLM provider API key
- Software
Node.js 22.22.1+@mcarvin/smart-diff (npm, run via npx)
Git
Permissions
Allowed Hosts
File Scopes
Reads git history only, read-only; it writes no files. Network access downloads the CLI from npm and sends the selected diff to the LLM provider you configure. Only one provider key is needed. The skill always runs with --redact-secrets, which masks likely credentials in the diff before sending it. Keys are read from environment variables and are never printed or requested in chat.