Dev tools

kotlin-footguns publishes 224 coding-agent skills mined from a shipping Kotlin app

September 16, 2026 at 12:25 AM PT

GitHub Open Graph card for maxrave-dev/kotlin-footguns

Image: GitHub

Why it mattersA Kotlin or Compose team using Claude Code, Cursor or Copilot gets a domain-specific rule set that reflects real production failures on desktop JVM and native bindings, instead of the model guessing from tutorials.

A GitHub repository called kotlin-footguns collects 224 agent skills for Kotlin, Compose Multiplatform and the desktop JVM in a single Markdown corpus, all extracted from the maintainer's shipping music application SimpMusic. The repository was created on 20 August 2026 and has reached 443 stars in 27 days.

What is in the repository

The repository holds 27,000 lines across 224 files, grouped into 14 sections labelled A to N. The sections cover native code on the desktop JVM, desktop packaging and CI, Multiplatform architecture, media playback engines, databases and SQL, Compose theming, Compose components, screens and navigation, reactive state, repository patterns, background work, remote API consumption, Kotlin utilities, and engineering methods. Two later batches labelled delta-6 and delta-7 add more recent lessons. Each skill file is a single failure mode with the fix and a verification command the agent is expected to run.

The stated method is that the lessons come from real failures during the production life of SimpMusic, the maintainer's cross-platform music app, which has been shipping to Android, Windows, macOS and Linux users since April 2023 and currently has 11,272 stars on GitHub. The repository README says the lessons are mined from that codebase rather than from documentation, and that later batches required executing every verification command against the source tree before shipping.

Who it is for and how it plugs in

The skills are formatted for Claude Code, Cursor, Codex, Copilot, Windsurf, Gemini and similar coding agents that support a public skill catalogue. Each skill sits in its own Markdown file with a Traps section that lists the specific ways the standard technique fails on desktop JVM or in Compose, followed by the verification commands. The files are also readable by a human as a reference for the same problems.

The corpus is under GPL-3.0, which matches the licence of the SimpMusic source it was mined from. A team pulling one skill into a proprietary project needs to weigh that licence choice: the skills are Markdown, not linked code, but the licence terms still travel with them.

What this fixes for an agent working on Kotlin

Kotlin, Compose Multiplatform and the desktop JVM are three of the areas where a coding agent guessing from public tutorials produces plausible code that runs on Android and breaks on macOS, or compiles on JVM and hangs on the packaging step. The trap for a maintainer is the loop where the model writes the same wrong pattern in a different shape every time. A rule set that captures the wrong patterns and lists the tests that catch them changes that loop into a check the agent runs before it commits.

The value depends on whether the failures a specific project sees overlap with SimpMusic's failures. A music app touches native audio bindings, low-level media playback, cross-platform packaging and background work heavily, and a team writing a business tool in Compose Multiplatform will find fewer of those cells relevant. Section-level adoption reads more honest than pulling every skill.

The bar this clears is that the traction number is real and the source is verifiable: 443 stars in 27 days on a domain-specific rule set, mined from a codebase with an 11,272-star track record, is a stronger signal than a general-purpose "agent skills" repository with the same star count. A team on Kotlin is one clone and one skill install away from testing whether the fit is there.

Source

maxrave-dev/kotlin-footguns on GitHub, 20 August 2026 onward. Source project: maxrave-dev/SimpMusic.

Source: GitHub

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

More from AI News

BankMCP puts a read-only MCP server in front of Enable Banking, so Claude Code and Cursor can query your own bank accounts

A one-person MIT-licensed project called BankMCP wraps Enable Banking's PSD2 API in a self-hosted MCP server, so an MCP client such as Claude Code, Cursor, ChatGPT or Ollama can read balances and transactions from 2,700 European banks with no third party holding the data.

Source: GitHubDev tools

Engrim ships a local SQLite memory store that survives switching between Claude Code, Cursor, Antigravity and Codex

Engrim is a local-first SQLite memory engine for coding agents that carries architectural decisions and project state across sessions and across models, from Claude Code to Cursor to Google Antigravity to Codex CLI.

Source: Hacker NewsDev tools

video-talkcraft is an Agent Skill that turns Claude Code and Codex into an explainer-video studio, 766 stars in 15 days

video-talkcraft takes a voiceover script and a recorded voice track, aligns them at the character level, and renders a Remotion explainer video with 108 pre-tuned motion cards, kinetic captions, and a camera system that refuses to hold a static frame.

Source: GitHubDev tools