Diagram diffs, a WeChat memory, and a skill that asks questions first

Fastest-growing this period

Ranked by GitHub star events in this reporting window, without subtracting unstars. Project briefs and the reading notes below are editorial additions.

tt-a1i/archify : Show your team exactly what changed in the architecture

JavaScript

Turns an agent's structured description of a system into interactive HTML and SVG architecture diagrams. It can compare diagram snapshots, useful for showing a team which components and connections changed between two versions of a design.

+143

mattpocock/skills : Make your AI ask the right questions before it writes code

Shell

Give your coding assistant a skill that asks detailed questions before implementation. Use it to clarify vague requirements before the agent builds the wrong thing.

+123

affaan-m/ECC : Give your coding agent a plan, tests and a memory

JavaScript

A collection of coding-agent workflows for planning, tests, review and remembering useful context. If your assistant jumps straight into edits or forgets earlier decisions, this gives you reusable routines to adapt to your project.

+116

DietrichGebert/ponytail : Stop your AI turning a small fix into a new framework

JavaScript

A set of instructions that asks AI coding agents to reuse existing code and question whether a new abstraction is needed. Its rules also explicitly retain validation and error handling, useful when your agent keeps making a small task unnecessarily complicated.

+109

deepseek-ai/deepseek-harness : Explore DeepSeek's building blocks for an extensible agent

TypeScript

DeepSeek's agent runtime built around replaceable plugins, with a web interface for running it. Useful for developers exploring how to compose an agent's capabilities; it is a developer preview, so expect interfaces to change as you experiment.

+97

Rion-Wu-tech/wechat-intelligence-hub : Find the replies and promises buried in your WeChat history

Python

Reads your authorized local WeChat records to find pending replies, promises and follow-up leads. Use it to check where a conversation left off; full history needs database access configured first, and daily reports can be exported as Markdown and HTML.

+85

microsoft/markitdown : Turn a pile of mixed documents into Markdown

Python

Converts documents such as PDFs and Office files into Markdown for text-processing tools. Useful when you want to feed a folder of mixed documents into search or an AI workflow while retaining headings, lists and other readable structure where supported.

+77

cathrynlavery/diagram-design : Make your diagrams match the article around them

HTML

A coding-agent skill for turning an explanation or existing diagram source into HTML and SVG diagrams. It can incorporate a publication's colors and fonts, useful when a diagram needs to fit an article or presentation rather than look like a separate tool's default export.

+74

blader/humanizer : Find the phrases that make your draft sound like AI

Python

An editing skill that looks for repetitive, inflated and formulaic AI writing and proposes rewrites. Useful for getting a second pass over a draft, especially when you supply an example of the voice you actually want to keep.

+74

pallets/flask : Build a small Python web app, then add what you need

Python

A Python web framework with a small starting point: define a route and a function to respond to requests. Choose your own libraries and add extensions as the application needs more features.

+67

What I would look at first

Reading notes updated 2026-09-08; the ranking above preserves its original reporting window.

Start with tt-a1i/archify if you ever have to explain a system redesign in a meeting. You feed it a structured description of the system and get an interactive HTML or SVG diagram, and its snapshot comparison shows exactly which components and connections changed between two versions. A sensible first try: describe the current setup and the proposed one, generate both diagrams, and put the diff on screen when you walk the team through the change, so nobody has to guess what moved.

Rion-Wu-tech/wechat-intelligence-hub solves a different itch: remembering what was actually said. With your authorization it reads your local WeChat chats and produces a daily report of pending replies, promises you made, business leads, and contacts worth reconnecting with. You can also just ask in plain language where a conversation left off. Full chat history needs local database access configured first. Once connected, generate a daily report for a familiar conversation and check it against the original messages. A single question can also be answered directly; it does not have to produce a report.

If your coding agent keeps building the wrong thing, mattpocock/skills is the lighter fix. It installs a skill that makes the agent ask detailed questions before touching code. Pair it mentally with DietrichGebert/ponytail, which instead constrains agents after they start working, pushing them to reuse existing code and keep validation in place. Try a small task with a vague requirement: see which questions the agent asks, then check whether it changed any unrelated code.