Audio
🔊 AI Memory That Learns From Mistakes
📦 alibaizhanov/mengram
Mengram is an open-source toolkit and cloud API that gives AI assistants a better memory. Instead of only saving facts, it can also store past events and step-by-step routines, then improve those routines when a task fails. This solves a common problem with AI agents: they forget user preferences, repeat mistakes, and lose context between conversations. Developers can connect it to Python, JavaScript, or API-based apps, import old chats or notes, and generate a profile that personalizes future responses. In plain terms, it helps AI tools remember what matters, learn from experience, and become more useful over time.
🔗 https://github.com/alibaizhanov/mengram
📰 https://news.ycombinator.com/item?id=47151177
📢 @hackernewsgithubprojects
📦 alibaizhanov/mengram
Mengram is an open-source toolkit and cloud API that gives AI assistants a better memory. Instead of only saving facts, it can also store past events and step-by-step routines, then improve those routines when a task fails. This solves a common problem with AI agents: they forget user preferences, repeat mistakes, and lose context between conversations. Developers can connect it to Python, JavaScript, or API-based apps, import old chats or notes, and generate a profile that personalizes future responses. In plain terms, it helps AI tools remember what matters, learn from experience, and become more useful over time.
🔗 https://github.com/alibaizhanov/mengram
📰 https://news.ycombinator.com/item?id=47151177
📢 @hackernewsgithubprojects
Audio
🔊 Build a Whiteboard Without Starting From Zero
📦 tldraw/tldraw
tldraw is an open-source toolkit for building online whiteboards and “infinite canvas” drawing spaces used for brainstorming, diagrams, and collaborative planning. Instead of building drawing tools, shapes, zooming, undo/redo, and live collaboration from scratch, developers can add tldraw to a web app and customize the experience. It powers the tldraw whiteboard app itself, so it is both a real product and a reusable foundation. The main benefit is speed and reliability: teams can launch visual editing features much faster, avoid many hard-to-build canvas details, and start from a polished base rather than reinventing everything.
🔗 https://github.com/tldraw/tldraw
📰 https://news.ycombinator.com/item?id=47151545
📢 @hackernewsgithubprojects
📦 tldraw/tldraw
tldraw is an open-source toolkit for building online whiteboards and “infinite canvas” drawing spaces used for brainstorming, diagrams, and collaborative planning. Instead of building drawing tools, shapes, zooming, undo/redo, and live collaboration from scratch, developers can add tldraw to a web app and customize the experience. It powers the tldraw whiteboard app itself, so it is both a real product and a reusable foundation. The main benefit is speed and reliability: teams can launch visual editing features much faster, avoid many hard-to-build canvas details, and start from a polished base rather than reinventing everything.
🔗 https://github.com/tldraw/tldraw
📰 https://news.ycombinator.com/item?id=47151545
📢 @hackernewsgithubprojects
Audio
🔊 Can Your AI Call Out Nonsense?
📦 petergpt/bullshit-benchmark
This repository is a benchmark for testing whether AI chatbots can recognize nonsense questions and push back instead of confidently making things up. It provides a curated set of intentionally bogus prompts (like mixing unrelated business, science, or technical terms), scripts to run many models against them, and a scoring system that grades whether each model clearly challenges the bad premise. The results can be published into a simple viewer and leaderboard for side-by-side comparison. It solves a real trust problem: models often sound helpful even when a question is incoherent. That makes this project useful for evaluating model reliability, safety tuning, and product choices.
🔗 https://github.com/petergpt/bullshit-benchmark
📰 https://news.ycombinator.com/item?id=47151505
📢 @hackernewsgithubprojects
📦 petergpt/bullshit-benchmark
This repository is a benchmark for testing whether AI chatbots can recognize nonsense questions and push back instead of confidently making things up. It provides a curated set of intentionally bogus prompts (like mixing unrelated business, science, or technical terms), scripts to run many models against them, and a scoring system that grades whether each model clearly challenges the bad premise. The results can be published into a simple viewer and leaderboard for side-by-side comparison. It solves a real trust problem: models often sound helpful even when a question is incoherent. That makes this project useful for evaluating model reliability, safety tuning, and product choices.
🔗 https://github.com/petergpt/bullshit-benchmark
📰 https://news.ycombinator.com/item?id=47151505
📢 @hackernewsgithubprojects
Audio
🔊 Turn Your Browser Into an AI-Powered Automation Assistant
📦 browserwing/browserwing
BrowserWing is an app and toolkit that helps people automate Chrome/Chromium through a local web interface, simple APIs, or AI assistants. Instead of repeating the same clicks, page visits, form filling, or data collection by hand, you can record a browser workflow once and replay it later. It also supports conversational commands and structured data extraction, which lowers the barrier for people who do not want to write much code. This is useful for repetitive research, website testing, page monitoring, and routine business tasks because it combines visual recording, browser control, and reusable integrations in one place.
🔗 https://github.com/browserwing/browserwing
📰 https://news.ycombinator.com/item?id=47151607
📢 @hackernewsgithubprojects
📦 browserwing/browserwing
BrowserWing is an app and toolkit that helps people automate Chrome/Chromium through a local web interface, simple APIs, or AI assistants. Instead of repeating the same clicks, page visits, form filling, or data collection by hand, you can record a browser workflow once and replay it later. It also supports conversational commands and structured data extraction, which lowers the barrier for people who do not want to write much code. This is useful for repetitive research, website testing, page monitoring, and routine business tasks because it combines visual recording, browser control, and reusable integrations in one place.
🔗 https://github.com/browserwing/browserwing
📰 https://news.ycombinator.com/item?id=47151607
📢 @hackernewsgithubprojects
Audio
🔊 AI Helpers for Airflow, in One Toolkit
📦 astronomer/agents
Astronomer’s agents repository is an open-source toolkit that adds practical AI help to data engineering work, especially for Apache Airflow (a popular system for scheduling data jobs). It combines ready-made “skills” for coding assistants, an Airflow connector/server, and a simple terminal tool so people can inspect pipelines, trigger runs, check logs, and work with data warehouses without jumping across many tools. The main problem it solves is setup friction: AI assistants often lack the right context or safe ways to interact with data systems. This repo gives them structured access and repeatable workflows, which makes debugging, building, and maintaining data pipelines faster and easier.
🔗 https://github.com/astronomer/agents
📰 https://news.ycombinator.com/item?id=47151588
📢 @hackernewsgithubprojects
📦 astronomer/agents
Astronomer’s agents repository is an open-source toolkit that adds practical AI help to data engineering work, especially for Apache Airflow (a popular system for scheduling data jobs). It combines ready-made “skills” for coding assistants, an Airflow connector/server, and a simple terminal tool so people can inspect pipelines, trigger runs, check logs, and work with data warehouses without jumping across many tools. The main problem it solves is setup friction: AI assistants often lack the right context or safe ways to interact with data systems. This repo gives them structured access and repeatable workflows, which makes debugging, building, and maintaining data pipelines faster and easier.
🔗 https://github.com/astronomer/agents
📰 https://news.ycombinator.com/item?id=47151588
📢 @hackernewsgithubprojects
Audio
🔊 One Dashboard to Run Your Django App
📦 yassi/dj-control-room
Django Control Room is an open-source add-on for Django (a popular Python web framework) that creates a single “control room” inside the site’s admin area. Instead of jumping between separate tools to check things like cache data, Redis, background jobs, or URL routes, teams can use one dashboard and add panels for the tasks they need. It solves the problem of scattered admin and maintenance tools, which can slow down troubleshooting and routine operations. The plug-in approach also makes it flexible: you can start with a few panels and expand later. It’s useful because it saves time, improves visibility, and reduces custom admin work.
🔗 https://github.com/yassi/dj-control-room
📰 https://news.ycombinator.com/item?id=47151995
📢 @hackernewsgithubprojects
📦 yassi/dj-control-room
Django Control Room is an open-source add-on for Django (a popular Python web framework) that creates a single “control room” inside the site’s admin area. Instead of jumping between separate tools to check things like cache data, Redis, background jobs, or URL routes, teams can use one dashboard and add panels for the tasks they need. It solves the problem of scattered admin and maintenance tools, which can slow down troubleshooting and routine operations. The plug-in approach also makes it flexible: you can start with a few panels and expand later. It’s useful because it saves time, improves visibility, and reduces custom admin work.
🔗 https://github.com/yassi/dj-control-room
📰 https://news.ycombinator.com/item?id=47151995
📢 @hackernewsgithubprojects
Audio
🔊 Build Your Own AI Workflow Engine
📦 strongdm/attractor
Attractor is a blueprint repository that explains how to build an AI-powered workflow system and coding assistant loop, rather than a finished app you install and run. It focuses on describing the architecture, data formats, and execution flow for complex tasks that need multiple steps, decisions, and retries. The problem it solves is the mess of one-off scripts and hard-to-maintain automation when teams try to connect AI tools together. By providing clear specifications and examples, it helps developers create systems that are easier to visualize, test, extend, and adapt to different AI providers or internal tools.
🔗 https://github.com/strongdm/attractor
📰 https://news.ycombinator.com/item?id=47152719
📢 @hackernewsgithubprojects
📦 strongdm/attractor
Attractor is a blueprint repository that explains how to build an AI-powered workflow system and coding assistant loop, rather than a finished app you install and run. It focuses on describing the architecture, data formats, and execution flow for complex tasks that need multiple steps, decisions, and retries. The problem it solves is the mess of one-off scripts and hard-to-maintain automation when teams try to connect AI tools together. By providing clear specifications and examples, it helps developers create systems that are easier to visualize, test, extend, and adapt to different AI providers or internal tools.
🔗 https://github.com/strongdm/attractor
📰 https://news.ycombinator.com/item?id=47152719
📢 @hackernewsgithubprojects
Audio
🔊 A Database That Also Runs Your App
📦 clockworklabs/spacetimedb
SpacetimeDB is an open-source project that combines a database and backend server into one system for building real-time applications. Instead of splitting data storage, server logic, and live updates across different tools, developers can keep the app’s rules close to the data and connect clients to a single platform. It helps solve a common problem: products like multiplayer games, chat apps, and collaborative tools are often hard to build, sync, and scale. By reducing moving parts and handling fast shared updates, SpacetimeDB can make development simpler, cut backend complexity, and help teams ship interactive software faster.
🔗 https://github.com/clockworklabs/SpacetimeDB
📰 https://news.ycombinator.com/item?id=47152859
📢 @hackernewsgithubprojects
📦 clockworklabs/spacetimedb
SpacetimeDB is an open-source project that combines a database and backend server into one system for building real-time applications. Instead of splitting data storage, server logic, and live updates across different tools, developers can keep the app’s rules close to the data and connect clients to a single platform. It helps solve a common problem: products like multiplayer games, chat apps, and collaborative tools are often hard to build, sync, and scale. By reducing moving parts and handling fast shared updates, SpacetimeDB can make development simpler, cut backend complexity, and help teams ship interactive software faster.
🔗 https://github.com/clockworklabs/SpacetimeDB
📰 https://news.ycombinator.com/item?id=47152859
📢 @hackernewsgithubprojects
Audio
🔊 Make Any Devices Feel Like They’re on the Same Network
📦 zerotier/zerotierone
This repository contains the code for ZeroTier One, software that creates secure virtual networks over the internet. It lets computers, phones, servers, and devices in different places connect as if they were plugged into the same local network, without the usual VPN setup headaches, router changes, or fragile workarounds. ZeroTier helps solve remote access and private networking for home labs, small businesses, distributed teams, and apps that need device-to-device communication. It is useful because it combines simple joining and management with end-to-end encryption and direct connections when possible, making private networking easier to deploy, scale, and maintain across many kinds of devices.
🔗 https://github.com/zerotier/ZeroTierOne
📰 https://news.ycombinator.com/item?id=47153012
📢 @hackernewsgithubprojects
📦 zerotier/zerotierone
This repository contains the code for ZeroTier One, software that creates secure virtual networks over the internet. It lets computers, phones, servers, and devices in different places connect as if they were plugged into the same local network, without the usual VPN setup headaches, router changes, or fragile workarounds. ZeroTier helps solve remote access and private networking for home labs, small businesses, distributed teams, and apps that need device-to-device communication. It is useful because it combines simple joining and management with end-to-end encryption and direct connections when possible, making private networking easier to deploy, scale, and maintain across many kinds of devices.
🔗 https://github.com/zerotier/ZeroTierOne
📰 https://news.ycombinator.com/item?id=47153012
📢 @hackernewsgithubprojects
Audio
🔊 Orca Turns LinkedIn Activity Into Clear Insight
📦 dimimikadze/orca
Orca is an open-source web app and reusable Node.js library that uses AI to analyze a LinkedIn profile in depth. Instead of relying only on a profile page or resume, it collects activity data such as posts, comments, reactions, and engagement patterns, then turns that information into structured insights you define, like interests, expertise, priorities, communication style, and how someone’s focus changes over time. This is useful for sales, recruiting, and investing because it helps people understand someone faster and prepare better outreach or decisions. It saves time by automating a research task that is usually slow, manual, and easy to miss details in.
🔗 https://github.com/DimiMikadze/orca
📰 https://news.ycombinator.com/item?id=47153615
📢 @hackernewsgithubprojects
📦 dimimikadze/orca
Orca is an open-source web app and reusable Node.js library that uses AI to analyze a LinkedIn profile in depth. Instead of relying only on a profile page or resume, it collects activity data such as posts, comments, reactions, and engagement patterns, then turns that information into structured insights you define, like interests, expertise, priorities, communication style, and how someone’s focus changes over time. This is useful for sales, recruiting, and investing because it helps people understand someone faster and prepare better outreach or decisions. It saves time by automating a research task that is usually slow, manual, and easy to miss details in.
🔗 https://github.com/DimiMikadze/orca
📰 https://news.ycombinator.com/item?id=47153615
📢 @hackernewsgithubprojects
Audio
🔊 Ask France’s Open Data Like a Chat
📦 datagouv/datagouv-mcp
datagouv-mcp is an open-source connector that lets AI chatbots talk directly to data.gouv.fr, France’s national open data website. Instead of manually searching pages, downloading files, and figuring out which dataset or API to use, a person can ask questions in plain language and the chatbot can search datasets, list files, pull dataset details, preview data rows, and inspect public APIs. It also supports usage metrics for datasets and resources. This is useful for journalists, students, researchers, and public servants because it makes public data faster to find and easier to explore through everyday chat tools like ChatGPT, Claude, VS Code, and others.
🔗 https://github.com/datagouv/datagouv-mcp
📰 https://news.ycombinator.com/item?id=47151805
📢 @hackernewsgithubprojects
📦 datagouv/datagouv-mcp
datagouv-mcp is an open-source connector that lets AI chatbots talk directly to data.gouv.fr, France’s national open data website. Instead of manually searching pages, downloading files, and figuring out which dataset or API to use, a person can ask questions in plain language and the chatbot can search datasets, list files, pull dataset details, preview data rows, and inspect public APIs. It also supports usage metrics for datasets and resources. This is useful for journalists, students, researchers, and public servants because it makes public data faster to find and easier to explore through everyday chat tools like ChatGPT, Claude, VS Code, and others.
🔗 https://github.com/datagouv/datagouv-mcp
📰 https://news.ycombinator.com/item?id=47151805
📢 @hackernewsgithubprojects
Audio
🔊 LadybugDB: Fast Graph Data, No Server Drama
📦 ladybugdb/ladybug
Ladybug is an open-source graph database, which means it stores information as connected items (people, products, events, relationships) and helps apps ask questions about those connections quickly. Unlike many databases that require a separate server, Ladybug can run directly inside your app or from a simple command-line tool, making setup easier and cheaper. It is designed for large datasets and heavy analysis, with built-in features for text search, vector search, and reliable transactions. This is useful for teams building data-rich products, recommendations, fraud checks, knowledge tools, or AI assistants that need fast answers from complex connected data.
🔗 https://github.com/LadybugDB/ladybug
📰 https://news.ycombinator.com/item?id=47153728
📢 @hackernewsgithubprojects
📦 ladybugdb/ladybug
Ladybug is an open-source graph database, which means it stores information as connected items (people, products, events, relationships) and helps apps ask questions about those connections quickly. Unlike many databases that require a separate server, Ladybug can run directly inside your app or from a simple command-line tool, making setup easier and cheaper. It is designed for large datasets and heavy analysis, with built-in features for text search, vector search, and reliable transactions. This is useful for teams building data-rich products, recommendations, fraud checks, knowledge tools, or AI assistants that need fast answers from complex connected data.
🔗 https://github.com/LadybugDB/ladybug
📰 https://news.ycombinator.com/item?id=47153728
📢 @hackernewsgithubprojects
Audio
🔊 A Fast Message Bridge for .NET Apps
📦 nats-io/nats.net
NATS .NET is an open-source C#/.NET library that lets apps connect to a NATS message server, a system used to pass information between services and devices. Instead of wiring every app directly to every other app, developers can send messages through NATS, which makes systems easier to scale, update, and keep running. This repository provides the .NET tools for common jobs like publish/subscribe messaging, streams, key-value storage, and service-style communication. It is useful for teams building modern apps because it helps different parts of a system talk reliably and quickly, from cloud services to edge devices.
🔗 https://github.com/nats-io/nats.net
📰 https://news.ycombinator.com/item?id=47154674
📢 @hackernewsgithubprojects
📦 nats-io/nats.net
NATS .NET is an open-source C#/.NET library that lets apps connect to a NATS message server, a system used to pass information between services and devices. Instead of wiring every app directly to every other app, developers can send messages through NATS, which makes systems easier to scale, update, and keep running. This repository provides the .NET tools for common jobs like publish/subscribe messaging, streams, key-value storage, and service-style communication. It is useful for teams building modern apps because it helps different parts of a system talk reliably and quickly, from cloud services to edge devices.
🔗 https://github.com/nats-io/nats.net
📰 https://news.ycombinator.com/item?id=47154674
📢 @hackernewsgithubprojects
Audio
🔊 Manim in Your Browser, No Setup
📦 maloyan/manim-web
manim-web is a browser-based playground for creating math and science-style animations with Manim, a popular animation tool. Instead of installing Python, video tools, and a complex setup on your computer, you can write code, render animations, and preview results directly in a web page. It includes a code editor, examples, and export options, which makes it easy to test ideas quickly and learn by experimenting. The main problem it solves is setup friction, which often stops people before they begin. It is useful for students, teachers, and curious creators who want visual explanations or demos without spending time configuring a local development environment first.
🔗 https://github.com/maloyan/manim-web
📰 https://news.ycombinator.com/item?id=47155375
📢 @hackernewsgithubprojects
📦 maloyan/manim-web
manim-web is a browser-based playground for creating math and science-style animations with Manim, a popular animation tool. Instead of installing Python, video tools, and a complex setup on your computer, you can write code, render animations, and preview results directly in a web page. It includes a code editor, examples, and export options, which makes it easy to test ideas quickly and learn by experimenting. The main problem it solves is setup friction, which often stops people before they begin. It is useful for students, teachers, and curious creators who want visual explanations or demos without spending time configuring a local development environment first.
🔗 https://github.com/maloyan/manim-web
📰 https://news.ycombinator.com/item?id=47155375
📢 @hackernewsgithubprojects
Audio
🔊 Woxi: Wolfram-Style Math Without the Bloat
📦 ad-si/woxi
Woxi is an open-source tool that runs a version of the Wolfram Language (the language used by Mathematica) with a focus on speed and easy access. It lets people write and execute math-heavy scripts from the command line, use it in Jupyter notebooks, or even try it in a web demo. It solves a common problem for learners, hobbyists, and developers who want Wolfram-style scripting without slow startup times or licensing friction. That makes it useful for quick experiments, teaching, automating calculations, and interactive notebook work when you want a lighter, faster setup.
🔗 https://github.com/ad-si/Woxi
📰 https://news.ycombinator.com/item?id=47155526
📢 @hackernewsgithubprojects
📦 ad-si/woxi
Woxi is an open-source tool that runs a version of the Wolfram Language (the language used by Mathematica) with a focus on speed and easy access. It lets people write and execute math-heavy scripts from the command line, use it in Jupyter notebooks, or even try it in a web demo. It solves a common problem for learners, hobbyists, and developers who want Wolfram-style scripting without slow startup times or licensing friction. That makes it useful for quick experiments, teaching, automating calculations, and interactive notebook work when you want a lighter, faster setup.
🔗 https://github.com/ad-si/Woxi
📰 https://news.ycombinator.com/item?id=47155526
📢 @hackernewsgithubprojects
Audio
🔊 Find Copy-Paste Code Fast
📦 mibk/dupl
Dupl is an open-source command-line tool that scans Go projects and finds duplicated code blocks (code clones). It helps solve a common problem in software projects: copy-pasted logic that spreads bugs, makes updates harder, and increases maintenance work because the same fix may need to be applied in multiple places. Dupl compares the structure of code, so it can spot repeats even when names or numbers are changed. That makes it useful for cleanup, refactoring, code reviews, and improving code quality. It can search whole folders, adjust how strict matching is, and output results in formats that are easy to read or process.
🔗 https://github.com/mibk/dupl
📰 https://news.ycombinator.com/item?id=47155460
📢 @hackernewsgithubprojects
📦 mibk/dupl
Dupl is an open-source command-line tool that scans Go projects and finds duplicated code blocks (code clones). It helps solve a common problem in software projects: copy-pasted logic that spreads bugs, makes updates harder, and increases maintenance work because the same fix may need to be applied in multiple places. Dupl compares the structure of code, so it can spot repeats even when names or numbers are changed. That makes it useful for cleanup, refactoring, code reviews, and improving code quality. It can search whole folders, adjust how strict matching is, and output results in formats that are easy to read or process.
🔗 https://github.com/mibk/dupl
📰 https://news.ycombinator.com/item?id=47155460
📢 @hackernewsgithubprojects
Audio
🔊 The Engine Behind the XRP Ledger
📦 xrplf/rippled
`rippled` is the open-source server software that powers the XRP Ledger, a shared network for payments and digital assets. It lets companies and independent operators run a node that connects to other nodes, checks transactions, and helps keep the ledger synchronized without depending on one central company or database. In simple terms, it solves the problem of keeping a trusted, up-to-date record of transfers, balances, and exchanges across a global network. Because the code is public and configurable, developers, exchanges, and infrastructure teams can build services on XRP Ledger, verify activity for themselves, and run payment or data systems with more transparency and control.
🔗 https://github.com/XRPLF/rippled
📰 https://news.ycombinator.com/item?id=47155753
📢 @hackernewsgithubprojects
📦 xrplf/rippled
`rippled` is the open-source server software that powers the XRP Ledger, a shared network for payments and digital assets. It lets companies and independent operators run a node that connects to other nodes, checks transactions, and helps keep the ledger synchronized without depending on one central company or database. In simple terms, it solves the problem of keeping a trusted, up-to-date record of transfers, balances, and exchanges across a global network. Because the code is public and configurable, developers, exchanges, and infrastructure teams can build services on XRP Ledger, verify activity for themselves, and run payment or data systems with more transparency and control.
🔗 https://github.com/XRPLF/rippled
📰 https://news.ycombinator.com/item?id=47155753
📢 @hackernewsgithubprojects
Audio
🔊 Put Claude Code on Autopilot
📦 jshchnz/claude-code-scheduler
Claude Code Scheduler is a plugin that lets people schedule Claude Code tasks in plain English, like “review my code every weekday at 9am” or “run a reminder tomorrow.” It solves the problem of having to remember repetitive checks, reviews, and maintenance jobs or keep Claude open to run them manually. The tool connects those requests to your computer’s built-in scheduling system, runs the task at the right time, and saves logs so you can see what happened later. It is useful because it turns routine coding work into reliable automation, helps teams catch issues earlier, and saves time for higher-value work.
🔗 https://github.com/jshchnz/claude-code-scheduler
📰 https://news.ycombinator.com/item?id=47155833
📢 @hackernewsgithubprojects
📦 jshchnz/claude-code-scheduler
Claude Code Scheduler is a plugin that lets people schedule Claude Code tasks in plain English, like “review my code every weekday at 9am” or “run a reminder tomorrow.” It solves the problem of having to remember repetitive checks, reviews, and maintenance jobs or keep Claude open to run them manually. The tool connects those requests to your computer’s built-in scheduling system, runs the task at the right time, and saves logs so you can see what happened later. It is useful because it turns routine coding work into reliable automation, helps teams catch issues earlier, and saves time for higher-value work.
🔗 https://github.com/jshchnz/claude-code-scheduler
📰 https://news.ycombinator.com/item?id=47155833
📢 @hackernewsgithubprojects
Audio
🔊 A Playbook for Smarter AI Agent Skills
📦 mgechev/skills-best-practices
This repository is a practical guide for people building “skills” (task-specific instruction packs) for AI agents. It explains how to organize a skill folder, write clear metadata so agents pick the right skill, keep instructions short to save context space, and move repetitive work into small scripts. It also includes a validation workflow: testing descriptions, simulating agent behavior, probing edge cases, and refining the skill structure. The problem it solves is unreliable or bloated agent skills that trigger at the wrong time or miss important steps. It is useful because it turns trial-and-error into a repeatable process for making agent skills more accurate and maintainable.
🔗 https://github.com/mgechev/skills-best-practices
📰 https://news.ycombinator.com/item?id=47155888
📢 @hackernewsgithubprojects
📦 mgechev/skills-best-practices
This repository is a practical guide for people building “skills” (task-specific instruction packs) for AI agents. It explains how to organize a skill folder, write clear metadata so agents pick the right skill, keep instructions short to save context space, and move repetitive work into small scripts. It also includes a validation workflow: testing descriptions, simulating agent behavior, probing edge cases, and refining the skill structure. The problem it solves is unreliable or bloated agent skills that trigger at the wrong time or miss important steps. It is useful because it turns trial-and-error into a repeatable process for making agent skills more accurate and maintainable.
🔗 https://github.com/mgechev/skills-best-practices
📰 https://news.ycombinator.com/item?id=47155888
📢 @hackernewsgithubprojects
Audio
🔊 Tree-sitter for Go, Minus the C Headaches
📦 odvcencio/gotreesitter
gotreesitter is a Go library that helps software understand source code, making it useful for editors, linters, search tools, and other developer apps. It re-creates the Tree-sitter parsing engine in pure Go, so teams do not need C libraries or a system compiler to build or run it. That solves a common setup and portability problem, especially when shipping tools across different environments or to WebAssembly. It can reuse existing language grammars and supports fast re-parsing after small edits, along with features like syntax highlighting, code queries, and symbol tagging. In short, it makes code-aware features easier to build and deploy in Go.
🔗 https://github.com/odvcencio/gotreesitter
📰 https://news.ycombinator.com/item?id=47155597
📢 @hackernewsgithubprojects
📦 odvcencio/gotreesitter
gotreesitter is a Go library that helps software understand source code, making it useful for editors, linters, search tools, and other developer apps. It re-creates the Tree-sitter parsing engine in pure Go, so teams do not need C libraries or a system compiler to build or run it. That solves a common setup and portability problem, especially when shipping tools across different environments or to WebAssembly. It can reuse existing language grammars and supports fast re-parsing after small edits, along with features like syntax highlighting, code queries, and symbol tagging. In short, it makes code-aware features easier to build and deploy in Go.
🔗 https://github.com/odvcencio/gotreesitter
📰 https://news.ycombinator.com/item?id=47155597
📢 @hackernewsgithubprojects